From 4e5b9f7107f042055581a6e9f53de6423007afa1 Mon Sep 17 00:00:00 2001
From: functionstackx <47992694+functionstackx@users.noreply.github.com>
Date: Sat, 26 Sep 2026 02:22:22 -0400
Subject: [PATCH 1/6] [Klaud Cold] Remove everything that pointed at the
 deleted benchmarks/single_node/agentic

reached it:
- Drop the 17 search-space entries without srt-recipe: six configs that were
  entirely bash-based (both B300 Qwen3.5 power A/B keys, qwen3.8next H100,
  Kimi-K3 and MiniMax-M3 MI355X vLLM, DSv4.1 Flash MI355X ATOM) and the ATOM
  LMCache DRAM-offload arms of Kimi-K3, MiniMax-M3 and GLM-5.2 on MI355X.
- Remove the dead dsv41flash bash branches in the GB200/GB300 launchers and
  the MiniMax-M3 ATOM script special case in the MI355X launcher.
- Restore the DSv4.1 Flash H100/H200 block32 installers and kernel configs
  under benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/
  and repoint the setup scripts, which the live single-node H100/H200 SGLang
  AgentX recipes run and #3461 broke.
- Point docs and comments at the srt-slurm recipes and srt_agentic.sh.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
---
 .github/codeowner-signoff-verify-prompt.md    |   2 +-
 KLAUD_DEBUG.md                                |   2 +-
 MODELS.md                                     |   2 +-
 MODELS_zh.md                                  |   2 +-
 .../multi_node/amd_utils/server_sglang.sh     |   4 +-
 .../install_h100_block32_configs.py           |  48 +++++++
 .../install_h200_block32_configs.py           |  43 ++++++
 ...3,dtype=fp8_w8a8,block_shape=[32, 32].json | 114 ++++++++++++++++
 ...3,dtype=fp8_w8a8,block_shape=[32, 32].json |  50 +++++++
 ...3,dtype=fp8_w8a8,block_shape=[32, 32].json |  98 ++++++++++++++
 ...3,dtype=fp8_w8a8,block_shape=[32, 32].json |  66 ++++++++++
 ...3,dtype=fp8_w8a8,block_shape=[32, 32].json |  66 ++++++++++
 ...3,dtype=fp8_w8a8,block_shape=[32, 32].json | 122 ++++++++++++++++++
 .../h100_dsv41_block32/provenance.json        |  71 ++++++++++
 ...0,dtype=fp8_w8a8,block_shape=[32, 32].json |  58 +++++++++
 ...0,dtype=fp8_w8a8,block_shape=[32, 32].json |  66 ++++++++++
 ...0,dtype=fp8_w8a8,block_shape=[32, 32].json |  66 ++++++++++
 ...0,dtype=fp8_w8a8,block_shape=[32, 32].json | 122 ++++++++++++++++++
 .../h200_dsv41_block32/provenance.json        |  16 +++
 .../configs/dsv41flash-h100-block32.sh        |   4 +-
 .../configs/dsv41flash-h200-block32.sh        |   4 +-
 .../glm5.2/atom/mi355x-fp4-mtp/agentic.yaml   |   4 +-
 .../kimik3/atom/mi355x-fp4-mtp/agentic.yaml   |   5 +-
 .../kimik3/vllm/mi355x-fp4-mtp/agentic.yaml   |   5 +-
 .../atom/mi355x-fp4-mtp/agentic.yaml          |   5 +-
 configs/amd-master.yaml                       |  55 --------
 configs/nvidia-master.yaml                    |  51 --------
 docs/DOCUMENTATION_PLAN.md                    |   4 +-
 docs/DOCUMENTATION_PLAN_zh.md                 |   4 +-
 docs/index.md                                 |   2 +-
 docs/index_zh.md                              |   2 +-
 docs/waiver/2508.md                           |   5 +-
 runners/launch_gb200-nv.sh                    |  35 -----
 runners/launch_gb300-nv.sh                    |  33 -----
 runners/launch_mi355x-amds.sh                 |   7 -
 35 files changed, 1036 insertions(+), 207 deletions(-)
 create mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/install_h100_block32_configs.py
 create mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/install_h200_block32_configs.py
 create mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=1280,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
 create mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=1792,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
 create mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=512,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
 create mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=5120,K=1024,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
 create mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=5120,K=288,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
 create mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=576,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
 create mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/provenance.json
 create mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=1792,K=5120,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json
 create mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=5120,K=1024,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json
 create mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=5120,K=288,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json
 create mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=576,K=5120,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json
 create mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/provenance.json

diff --git a/.github/codeowner-signoff-verify-prompt.md b/.github/codeowner-signoff-verify-prompt.md
index e2780d4217..d29eaf8bd7 100644
--- a/.github/codeowner-signoff-verify-prompt.md
+++ b/.github/codeowner-signoff-verify-prompt.md
@@ -316,7 +316,7 @@ The rule covers the SERVING stack that produces the numbers.
 ## Check 11 — Agentic spec-decode configs use the golden simulated acceptance length
 APPLICABILITY: this check covers AGENTIC-workload benchmark changes that enable
 speculative decoding. From the PR diff, identify configs that are BOTH:
-- agentic scripts under `benchmarks/single_node/agentic/**`, multi-node recipes
+- single-node agentic srt-slurm recipes (`benchmarks/single_node/srt-slurm-recipes/**/agentic.yaml`), multi-node recipes
   under an `agentic/` directory (e.g. `benchmarks/multi_node/srt-slurm-recipes/**/agentic/**`),
   or master-config entries whose name/recipe path marks them agentic, AND
 - speculative-decoding with MTP / EAGLE / draft-model flags such as
diff --git a/KLAUD_DEBUG.md b/KLAUD_DEBUG.md
index 76be01cb97..0f0fb21043 100644
--- a/KLAUD_DEBUG.md
+++ b/KLAUD_DEBUG.md
@@ -77,7 +77,7 @@ changing the benchmark recipe.
 
 **Fix:** in `benchmarks/single_node/<recipe>.sh`, either:
 1. **Lower `--gpu-memory-utilization`** (`0.95 → 0.90`, sometimes 0.85). Matches the H100/H200/B200 NVIDIA pattern. Smallest blast radius.
-2. **Disable the profiler entirely** for cases where lowering isn't enough: `export VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS=0` before `vllm serve`. Matches `benchmarks/single_node/agentic/deprecated/kimik2.5_fp4_b200.sh:65`.
+2. **Disable the profiler entirely** for cases where lowering isn't enough: `export VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS=0` before `vllm serve`. Matches `benchmarks/single_node/fixed_seq_len/deprecated/kimik2.5_fp4_b200.sh`.
 
 Seen on: #1395 (kimik2.5-fp4-b200-vllm, needed env var), #1403 (gptoss-fp4-mi300x-vllm, needed 0.90), #1461 (dsv4-fp8-h200-vllm, needed 0.90).
 
diff --git a/MODELS.md b/MODELS.md
index c9fc1adc3b..ce8e6f83f0 100644
--- a/MODELS.md
+++ b/MODELS.md
@@ -64,7 +64,7 @@ Rationale: `dsv4` carries the largest single-turn footprint in the repository. 4
 
 | Scenario | ISL/OSL | Status |
 |---|---|---|
-| Agentic coding | Long Context, Multi Turn Realistic traffic trace replay with sub agents | Active. This is the trace-replay agentic-coding benchmark (see [`benchmarks/single_node/agentic/`](benchmarks/single_node/agentic/)). Going forward, new models will likely be onboarded with agentic coding only. Speculative decoding may be enabled or disabled to produce the best Pareto points; a separate non-spec-decode A/B baseline is not required (see [Deprecation Notice](#deprecation-notice)). |
+| Agentic coding | Long Context, Multi Turn Realistic traffic trace replay with sub agents | Active. This is the trace-replay agentic-coding benchmark (see the srt-slurm recipes under [`benchmarks/single_node/srt-slurm-recipes/`](benchmarks/single_node/srt-slurm-recipes/) and the shared client [`benchmarks/srt_agentic.sh`](benchmarks/srt_agentic.sh)). Going forward, new models will likely be onboarded with agentic coding only. Speculative decoding may be enabled or disabled to produce the best Pareto points; a separate non-spec-decode A/B baseline is not required (see [Deprecation Notice](#deprecation-notice)). |
 | Single-turn 8k1k | 8192 / 1024 | Active. This is the primary fixed-sequence-length scenario. |
 | Single-turn 1k1k | 1024 / 1024 | Deprecated since 2026-07-17 ([#2263](https://github.com/SemiAnalysisAI/InferenceX/pull/2263)), to save GPU cluster time for higher-priority real-world agentic-coding benchmarks and new frontier models. Archived configs live in [`configs/deprecated/`](configs/deprecated/). The GLM-5.1 B200 TileRT point added later in [#2533](https://github.com/SemiAnalysisAI/InferenceX/pull/2533) remains active. |
 | Single-turn 1k8k | 1024 / 8192 | **Deprecated for all models** since 2026-03-27 ([#911](https://github.com/SemiAnalysisAI/InferenceX/pull/911)), to save GPU cluster time for higher-priority real-world agentic-coding benchmarks and new frontier models. Configs were removed, not archived. |
diff --git a/MODELS_zh.md b/MODELS_zh.md
index f26bc9b947..e2794deb5b 100644
--- a/MODELS_zh.md
+++ b/MODELS_zh.md
@@ -64,7 +64,7 @@ InferenceX-e2e 运行在数量固定且有限的 GPU 资源池上，并由一支
 
 | 场景 | ISL/OSL | 状态 |
 |---|---|---|
-| 智能体编码（agentic coding） | 长上下文、多轮真实流量的轨迹回放，含子智能体（sub agents） | 启用。此场景采用基于轨迹回放的智能体编码基准测试（见 [`benchmarks/single_node/agentic/`](benchmarks/single_node/agentic/)）。今后新模型预计将仅以智能体编码场景接入。可开启或关闭投机解码以获得最优帕累托点；不要求独立的非投机解码 A/B 基线（见[弃用公告](#弃用公告)）。 |
+| 智能体编码（agentic coding） | 长上下文、多轮真实流量的轨迹回放，含子智能体（sub agents） | 启用。此场景采用基于轨迹回放的智能体编码基准测试（见 [`benchmarks/single_node/srt-slurm-recipes/`](benchmarks/single_node/srt-slurm-recipes/) 下的 srt-slurm 配方与共享客户端 [`benchmarks/srt_agentic.sh`](benchmarks/srt_agentic.sh)）。今后新模型预计将仅以智能体编码场景接入。可开启或关闭投机解码以获得最优帕累托点；不要求独立的非投机解码 A/B 基线（见[弃用公告](#弃用公告)）。 |
 | 单轮 8k1k | 8192 / 1024 | 启用。当前主要的固定序列长度（fixed-seq-len）场景。 |
 | 单轮 1k1k | 1024 / 1024 | 自 2026-07-17 起弃用（[#2263](https://github.com/SemiAnalysisAI/InferenceX/pull/2263)），以便将 GPU 集群时间留给优先级更高的真实场景智能体编码基准测试与新的前沿模型。归档配置位于 [`configs/deprecated/`](configs/deprecated/)。后续由 [#2533](https://github.com/SemiAnalysisAI/InferenceX/pull/2533) 加入的 GLM-5.1 B200 TileRT 测试点仍启用。 |
 | 单轮 1k8k | 1024 / 8192 | **对所有模型均已弃用**，自 2026-03-27 起（[#911](https://github.com/SemiAnalysisAI/InferenceX/pull/911)），以便将 GPU 集群时间留给优先级更高的真实场景智能体编码基准测试与新的前沿模型。相关配置已删除，未归档。 |
diff --git a/benchmarks/multi_node/amd_utils/server_sglang.sh b/benchmarks/multi_node/amd_utils/server_sglang.sh
index cae7c9d25f..ffb18695e9 100755
--- a/benchmarks/multi_node/amd_utils/server_sglang.sh
+++ b/benchmarks/multi_node/amd_utils/server_sglang.sh
@@ -580,8 +580,8 @@ elif [[ "$KV_OFFLOADING" != "none" && "$KV_OFFLOAD_BACKEND" == umbp-linker* ]];
     # =========================================================================
     # UMBP as a DIRECT external store for the unified radix tree (PD disagg).
     #
-    # Ported from benchmarks/single_node/agentic/dsv4_fp4_mi355x_sglang_mtp.sh,
-    # which is where this arm exists today. It is a SEPARATE sglang code path
+    # Ported from the single-node DSv4 MI355X SGLang AgentX bash script (since
+    # removed in #3460). It is a SEPARATE sglang code path
     # from the HiCache branch above, not a variation of it: the tree loads and
     # offloads pages against UMBP with NO host cache tier in between, and
     # sglang rejects the combination outright (server_args.py::_handle_hicache
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/install_h100_block32_configs.py b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/install_h100_block32_configs.py
new file mode 100644
index 0000000000..88c3fc97f6
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/install_h100_block32_configs.py
@@ -0,0 +1,48 @@
+"""Install the H100 V4.1 tiling configs into the imported SGLang kernel package."""  # noqa: INP001 -- Standalone recipe executable.
+
+import hashlib
+import json
+import shutil
+import sys
+from pathlib import Path
+
+from sglang.kernels.ops.quantization import fp8_kernel
+
+
+def main() -> None:
+    source = Path(sys.argv[1])
+    artifacts = Path(sys.argv[2]) / "fp8_kernel_configs"
+    device = fp8_kernel.get_device_name().replace(" ", "_")
+    if device != "NVIDIA_H100_80GB_HBM3":
+        raise RuntimeError(f"H100 tiling configs cannot be installed on {device}")
+    destination = Path(fp8_kernel.__file__).resolve().parent / "configs"
+    destination.mkdir(exist_ok=True)
+    artifacts.mkdir(parents=True, exist_ok=True)
+    for n, k in [
+        (5120, 1024),
+        (5120, 288),
+        (576, 5120),
+        (1792, 5120),
+        (512, 5120),
+        (1280, 5120),
+    ]:
+        name = (
+            f"N={n},K={k},device_name={device},dtype=fp8_w8a8,block_shape=[32, 32].json"
+        )
+        path = source / name
+        expected = {
+            int(m): config for m, config in json.loads(path.read_text()).items()
+        }
+        shutil.copyfile(path, destination / name)
+        shutil.copyfile(path, artifacts / name)
+        fp8_kernel.get_w8a8_block_fp8_configs.cache_clear()
+        actual = fp8_kernel.get_w8a8_block_fp8_configs(n, k, 32, 32)
+        if actual != expected:
+            raise RuntimeError(f"SGLang did not resolve the installed config: {name}")
+        print(
+            f"Installed {destination / name}: sha256={hashlib.sha256(path.read_bytes()).hexdigest()}"
+        )
+
+
+if __name__ == "__main__":
+    main()
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/install_h200_block32_configs.py b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/install_h200_block32_configs.py
new file mode 100644
index 0000000000..3d34a865e2
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/install_h200_block32_configs.py
@@ -0,0 +1,43 @@
+"""Install the H200 V4.1 tiling configs into the imported SGLang kernel package."""  # noqa: INP001 -- Standalone recipe executable.
+
+import hashlib
+import json
+import shutil
+import sys
+from pathlib import Path
+
+from sglang.kernels.ops.quantization import fp8_kernel
+
+
+def main() -> None:
+    source = Path(sys.argv[1])
+    artifacts = Path(sys.argv[2]) / "fp8_kernel_configs"
+    tp = int(sys.argv[3])
+    if tp not in (4, 8):
+        raise ValueError(f"H200 tiling configs support TP4 or TP8, got TP{tp}")
+    device = fp8_kernel.get_device_name().replace(" ", "_")
+    if device != "NVIDIA_H200":
+        raise RuntimeError(f"H200 tiling configs cannot be installed on {device}")
+    destination = Path(fp8_kernel.__file__).resolve().parent / "configs"
+    destination.mkdir(exist_ok=True)
+    artifacts.mkdir(parents=True, exist_ok=True)
+    shapes = [(1792, 5120)]
+    if tp == 8:
+        shapes += [(5120, 1024), (5120, 288), (576, 5120)]
+    for n, k in shapes:
+        name = f"N={n},K={k},device_name={device},dtype=fp8_w8a8,block_shape=[32, 32].json"
+        path = source / name
+        expected = {int(m): config for m, config in json.loads(path.read_text()).items()}
+        shutil.copyfile(path, destination / name)
+        shutil.copyfile(path, artifacts / name)
+        fp8_kernel.get_w8a8_block_fp8_configs.cache_clear()
+        actual = fp8_kernel.get_w8a8_block_fp8_configs(n, k, 32, 32)
+        if actual != expected:
+            raise RuntimeError(f"SGLang did not resolve the installed config: {name}")
+        print(
+            f"Installed {destination / name}: sha256={hashlib.sha256(path.read_bytes()).hexdigest()}"
+        )
+
+
+if __name__ == "__main__":
+    main()
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=1280,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=1280,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
new file mode 100644
index 0000000000..971110754f
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=1280,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json	
@@ -0,0 +1,114 @@
+{
+  "0": {
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3
+  },
+  "1": {
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3,
+    "SPLIT_K": 16,
+    "SWAP_AB": false
+  },
+  "5": {
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3,
+    "SPLIT_K": 16,
+    "SWAP_AB": false
+  },
+  "10": {
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3,
+    "SPLIT_K": 16,
+    "SWAP_AB": true
+  },
+  "20": {
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3,
+    "SPLIT_K": 8,
+    "SWAP_AB": true
+  },
+  "40": {
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3,
+    "SPLIT_K": 8,
+    "SWAP_AB": true
+  },
+  "80": {
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3,
+    "SPLIT_K": 16,
+    "SWAP_AB": true
+  },
+  "160": {
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3,
+    "SPLIT_K": 8,
+    "SWAP_AB": true
+  },
+  "190": {
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3,
+    "SPLIT_K": 8,
+    "SWAP_AB": true
+  },
+  "191": {
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3
+  },
+  "320": {
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3
+  },
+  "321": {
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3
+  }
+}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=1792,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=1792,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
new file mode 100644
index 0000000000..986ff1953b
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=1792,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json	
@@ -0,0 +1,50 @@
+{
+  "1": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 16,
+    "SWAP_AB": false,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "6": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 16,
+    "SWAP_AB": true,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "24": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 8,
+    "SWAP_AB": true,
+    "num_stages": 4,
+    "num_warps": 4
+  },
+  "96": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 8,
+    "SWAP_AB": true,
+    "num_stages": 4,
+    "num_warps": 4
+  },
+  "128": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "num_stages": 3,
+    "num_warps": 4
+  }
+}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=512,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=512,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
new file mode 100644
index 0000000000..3c4f7326b0
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=512,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json	
@@ -0,0 +1,98 @@
+{
+  "0": {
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3
+  },
+  "1": {
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3,
+    "SPLIT_K": 16,
+    "SWAP_AB": false
+  },
+  "5": {
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3,
+    "SPLIT_K": 16,
+    "SWAP_AB": false
+  },
+  "10": {
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3,
+    "SPLIT_K": 16,
+    "SWAP_AB": false
+  },
+  "20": {
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3,
+    "SPLIT_K": 16,
+    "SWAP_AB": true
+  },
+  "40": {
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3,
+    "SPLIT_K": 16,
+    "SWAP_AB": true
+  },
+  "80": {
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3,
+    "SPLIT_K": 8,
+    "SWAP_AB": true
+  },
+  "160": {
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3,
+    "SPLIT_K": 4,
+    "SWAP_AB": true
+  },
+  "320": {
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3,
+    "SPLIT_K": 4,
+    "SWAP_AB": true
+  },
+  "321": {
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 32,
+    "BLOCK_SIZE_K": 32,
+    "GROUP_SIZE_M": 32,
+    "num_warps": 4,
+    "num_stages": 3
+  }
+}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=5120,K=1024,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=5120,K=1024,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
new file mode 100644
index 0000000000..fece264886
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=5120,K=1024,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json	
@@ -0,0 +1,66 @@
+{
+    "1": {
+        "BLOCK_SIZE_M": 16,
+        "BLOCK_SIZE_N": 64,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 16,
+        "num_warps": 4,
+        "num_stages": 5
+    },
+    "4": {
+        "BLOCK_SIZE_M": 16,
+        "BLOCK_SIZE_N": 64,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 64,
+        "num_warps": 4,
+        "num_stages": 4
+    },
+    "16": {
+        "BLOCK_SIZE_M": 16,
+        "BLOCK_SIZE_N": 32,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 1,
+        "num_warps": 4,
+        "num_stages": 4
+    },
+    "32": {
+        "BLOCK_SIZE_M": 16,
+        "BLOCK_SIZE_N": 64,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 16,
+        "num_warps": 4,
+        "num_stages": 5
+    },
+    "64": {
+        "BLOCK_SIZE_M": 64,
+        "BLOCK_SIZE_N": 128,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 1,
+        "num_warps": 4,
+        "num_stages": 5
+    },
+    "256": {
+        "BLOCK_SIZE_M": 64,
+        "BLOCK_SIZE_N": 128,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 16,
+        "num_warps": 4,
+        "num_stages": 5
+    },
+    "1024": {
+        "BLOCK_SIZE_M": 64,
+        "BLOCK_SIZE_N": 128,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 1,
+        "num_warps": 4,
+        "num_stages": 5
+    },
+    "4096": {
+        "BLOCK_SIZE_M": 64,
+        "BLOCK_SIZE_N": 128,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 1,
+        "num_warps": 4,
+        "num_stages": 4
+    }
+}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=5120,K=288,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=5120,K=288,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
new file mode 100644
index 0000000000..ccf51c6e0a
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=5120,K=288,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json	
@@ -0,0 +1,66 @@
+{
+    "1": {
+        "BLOCK_SIZE_M": 16,
+        "BLOCK_SIZE_N": 64,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 16,
+        "num_warps": 4,
+        "num_stages": 2
+    },
+    "4": {
+        "BLOCK_SIZE_M": 16,
+        "BLOCK_SIZE_N": 128,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 32,
+        "num_warps": 4,
+        "num_stages": 2
+    },
+    "16": {
+        "BLOCK_SIZE_M": 16,
+        "BLOCK_SIZE_N": 64,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 32,
+        "num_warps": 8,
+        "num_stages": 2
+    },
+    "32": {
+        "BLOCK_SIZE_M": 32,
+        "BLOCK_SIZE_N": 32,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 64,
+        "num_warps": 4,
+        "num_stages": 5
+    },
+    "64": {
+        "BLOCK_SIZE_M": 64,
+        "BLOCK_SIZE_N": 32,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 32,
+        "num_warps": 4,
+        "num_stages": 3
+    },
+    "256": {
+        "BLOCK_SIZE_M": 64,
+        "BLOCK_SIZE_N": 128,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 64,
+        "num_warps": 4,
+        "num_stages": 5
+    },
+    "1024": {
+        "BLOCK_SIZE_M": 64,
+        "BLOCK_SIZE_N": 128,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 16,
+        "num_warps": 4,
+        "num_stages": 2
+    },
+    "4096": {
+        "BLOCK_SIZE_M": 64,
+        "BLOCK_SIZE_N": 128,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 16,
+        "num_warps": 4,
+        "num_stages": 3
+    }
+}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=576,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=576,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
new file mode 100644
index 0000000000..4e45c66e6d
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=576,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json	
@@ -0,0 +1,122 @@
+{
+  "1": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 16,
+    "SWAP_AB": false,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "4": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 16,
+    "SWAP_AB": true,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "6": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 16,
+    "SWAP_AB": true,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "16": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 16,
+    "SWAP_AB": true,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "24": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 16,
+    "SWAP_AB": true,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "32": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 16,
+    "SWAP_AB": true,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "64": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 8,
+    "SWAP_AB": true,
+    "num_stages": 4,
+    "num_warps": 4
+  },
+  "96": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 8,
+    "SWAP_AB": true,
+    "num_stages": 4,
+    "num_warps": 4
+  },
+  "128": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 16,
+    "num_stages": 5,
+    "num_warps": 4
+  },
+  "193": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 32,
+    "num_stages": 4,
+    "num_warps": 4
+  },
+  "256": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 32,
+    "num_stages": 4,
+    "num_warps": 4
+  },
+  "1024": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 128,
+    "GROUP_SIZE_M": 1,
+    "num_stages": 4,
+    "num_warps": 4
+  },
+  "4096": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 128,
+    "GROUP_SIZE_M": 1,
+    "num_stages": 5,
+    "num_warps": 4
+  }
+}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/provenance.json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/provenance.json
new file mode 100644
index 0000000000..9c4fff096c
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/provenance.json
@@ -0,0 +1,71 @@
+{
+  "sglang_commit": "0f6761b54facebb47f2068f87ecccd8f14da3a0e",
+  "image": "lmsysorg/sglang:nightly-dev-cu13-20260921-0f6761b5",
+  "hardware": "NVIDIA H100 80GB HBM3",
+  "tuner": "benchmark/kernels/quantization/tuning_block_wise_kernel.py",
+  "search_change": "Use BLOCK_SIZE_K=[32] for checkpoint block32 scales; upstream CUDA candidates [64,128] otherwise filter to an empty search.",
+  "tuning_slurm_job": 18452,
+  "selected_audit_slurm_job": 18457,
+  "resolver_install_slurm_job": 18458,
+  "validation": "39 cases: FP32 dequantized reference rtol=0.02/atol=0.05; exact CUDA graph/eager equality; alternating timing order. BF16 rounding may differ from default tiling.",
+  "selection": "N5120/K288 batch64 restored to upstream default after timing regression.",
+  "precision": "Stock FP8 inputs and scales, BF16 output; no quantization or model changes.",
+  "split_k_audit": {
+    "slurm_job": 18479,
+    "cases": 80,
+    "mechanism": "Pinned upstream Hopper SPLIT_K and SWAP_AB launch configuration, FP32 partial reduction, unchanged FP8 inputs/scales and BF16 output",
+    "shapes": [
+      "N1792/K5120",
+      "N576/K5120"
+    ],
+    "boundary_validation": "Slurm18480:56/56 FP32-reference and exact CUDA-graph replay checks passed, batch boundariesM1..1024; changed cases all faster in alternating-order microtimings",
+    "large_batch_policy": "Retain original fallback selection for M129..8192 using explicit boundary anchors"
+  },
+  "draft_projection_audit": {
+    "device": "NVIDIA H100 80GB HBM3",
+    "image": "nightly-dev-cu13-20260922-582389ce",
+    "search": {
+      "job": 18492,
+      "cases": 144,
+      "status": "COMPLETED0:0"
+    },
+    "boundary": {
+      "job": 18493,
+      "cases": 206,
+      "status": "COMPLETED0:0"
+    },
+    "trimmed_boundary": {
+      "job": 18494,
+      "cases": 20,
+      "status": "COMPLETED0:0"
+    },
+    "reference": "FP32 dequant reference rtol.02/atol.05; exact eager/graph replay; no quantization or kernel code changes",
+    "ranges": {
+      "N512": "M1..320 tuned; M0 and M>=321 stock",
+      "N1280": "M1..190 tuned; M0 and M>=191 stock"
+    },
+    "retained_speedup_range": [
+      1.1861693666195017,
+      6.770688638312714
+    ],
+    "baseline_l2_max": 0.0016977186314761639,
+    "candidate_l2_max": 0.0016970767173916101,
+    "hashes": {
+      "N=1280,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json": "635f684b6ed8b538709249baa5c474bf8583f952bcee330e01fd0f9e0ce714a3",
+      "N=512,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json": "193fc012550dce39bb4de0a4bde28eed5cee79f2900ed95841fe18ccb6d8c187"
+    },
+    "serving_qualification": {
+      "candidate_run": 35683668032,
+      "baseline_run": 35683669703,
+      "protocol": "Matched canonical C1, 3600 seconds, normal warmup; zero profiled errors and valid power in both runs",
+      "candidate_total_tps_per_gpu": 1239.57314,
+      "baseline_total_tps_per_gpu": 958.51774,
+      "candidate_p90_interactivity": 90.89516,
+      "baseline_p90_interactivity": 60.93631,
+      "full_gsm8k_run": 35683671358,
+      "full_gsm8k_examples": 1319,
+      "full_gsm8k_strict_accuracy": 0.9696739954510993,
+      "final_sweep": "Required after integration"
+    }
+  }
+}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=1792,K=5120,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=1792,K=5120,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json
new file mode 100644
index 0000000000..83620a6b10
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=1792,K=5120,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json	
@@ -0,0 +1,58 @@
+{
+  "1": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 8,
+    "SWAP_AB": true,
+    "num_stages": 4,
+    "num_warps": 4
+  },
+  "2": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "6": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 16,
+    "SWAP_AB": true,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "24": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 8,
+    "SWAP_AB": true,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "96": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 16,
+    "SWAP_AB": true,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "128": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "num_stages": 3,
+    "num_warps": 4
+  }
+}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=5120,K=1024,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=5120,K=1024,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json
new file mode 100644
index 0000000000..850f61947e
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=5120,K=1024,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json	
@@ -0,0 +1,66 @@
+{
+    "1": {
+        "BLOCK_SIZE_M": 16,
+        "BLOCK_SIZE_N": 64,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 64,
+        "num_warps": 4,
+        "num_stages": 3
+    },
+    "4": {
+        "BLOCK_SIZE_M": 16,
+        "BLOCK_SIZE_N": 32,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 64,
+        "num_warps": 4,
+        "num_stages": 4
+    },
+    "16": {
+        "BLOCK_SIZE_M": 16,
+        "BLOCK_SIZE_N": 64,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 16,
+        "num_warps": 4,
+        "num_stages": 5
+    },
+    "32": {
+        "BLOCK_SIZE_M": 32,
+        "BLOCK_SIZE_N": 64,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 32,
+        "num_warps": 4,
+        "num_stages": 4
+    },
+    "64": {
+        "BLOCK_SIZE_M": 32,
+        "BLOCK_SIZE_N": 128,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 16,
+        "num_warps": 4,
+        "num_stages": 3
+    },
+    "256": {
+        "BLOCK_SIZE_M": 64,
+        "BLOCK_SIZE_N": 128,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 16,
+        "num_warps": 4,
+        "num_stages": 5
+    },
+    "1024": {
+        "BLOCK_SIZE_M": 64,
+        "BLOCK_SIZE_N": 128,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 32,
+        "num_warps": 4,
+        "num_stages": 3
+    },
+    "4096": {
+        "BLOCK_SIZE_M": 64,
+        "BLOCK_SIZE_N": 128,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 1,
+        "num_warps": 4,
+        "num_stages": 4
+    }
+}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=5120,K=288,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=5120,K=288,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json
new file mode 100644
index 0000000000..94b8a3fccf
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=5120,K=288,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json	
@@ -0,0 +1,66 @@
+{
+    "1": {
+        "BLOCK_SIZE_M": 32,
+        "BLOCK_SIZE_N": 64,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 32,
+        "num_warps": 8,
+        "num_stages": 2
+    },
+    "4": {
+        "BLOCK_SIZE_M": 64,
+        "BLOCK_SIZE_N": 128,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 1,
+        "num_warps": 4,
+        "num_stages": 4
+    },
+    "16": {
+        "BLOCK_SIZE_M": 32,
+        "BLOCK_SIZE_N": 128,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 64,
+        "num_warps": 4,
+        "num_stages": 2
+    },
+    "32": {
+        "BLOCK_SIZE_M": 64,
+        "BLOCK_SIZE_N": 32,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 32,
+        "num_warps": 4,
+        "num_stages": 3
+    },
+    "64": {
+        "BLOCK_SIZE_M": 16,
+        "BLOCK_SIZE_N": 128,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 16,
+        "num_warps": 4,
+        "num_stages": 4
+    },
+    "256": {
+        "BLOCK_SIZE_M": 64,
+        "BLOCK_SIZE_N": 128,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 64,
+        "num_warps": 4,
+        "num_stages": 3
+    },
+    "1024": {
+        "BLOCK_SIZE_M": 64,
+        "BLOCK_SIZE_N": 128,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 32,
+        "num_warps": 4,
+        "num_stages": 4
+    },
+    "4096": {
+        "BLOCK_SIZE_M": 64,
+        "BLOCK_SIZE_N": 128,
+        "BLOCK_SIZE_K": 32,
+        "GROUP_SIZE_M": 64,
+        "num_warps": 4,
+        "num_stages": 4
+    }
+}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=576,K=5120,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=576,K=5120,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json
new file mode 100644
index 0000000000..4359ec7f93
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=576,K=5120,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json	
@@ -0,0 +1,122 @@
+{
+  "1": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 16,
+    "SWAP_AB": false,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "4": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 16,
+    "SWAP_AB": false,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "6": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 16,
+    "SWAP_AB": false,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "16": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 16,
+    "SWAP_AB": true,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "24": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 16,
+    "SWAP_AB": true,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "32": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 16,
+    "SWAP_AB": true,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "64": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 8,
+    "SWAP_AB": true,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "96": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 32,
+    "GROUP_SIZE_M": 32,
+    "SPLIT_K": 8,
+    "SWAP_AB": true,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "128": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 16,
+    "num_stages": 5,
+    "num_warps": 4
+  },
+  "193": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 64,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "256": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 16,
+    "BLOCK_SIZE_N": 64,
+    "GROUP_SIZE_M": 64,
+    "num_stages": 3,
+    "num_warps": 4
+  },
+  "1024": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 128,
+    "GROUP_SIZE_M": 1,
+    "num_stages": 5,
+    "num_warps": 4
+  },
+  "4096": {
+    "BLOCK_SIZE_K": 32,
+    "BLOCK_SIZE_M": 64,
+    "BLOCK_SIZE_N": 128,
+    "GROUP_SIZE_M": 1,
+    "num_stages": 5,
+    "num_warps": 4
+  }
+}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/provenance.json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/provenance.json
new file mode 100644
index 0000000000..074329f645
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/provenance.json
@@ -0,0 +1,16 @@
+{
+  "sglang_commit": "0f6761b54facebb47f2068f87ecccd8f14da3a0e",
+  "image": "lmsysorg/sglang:nightly-dev-cu13-20260921-0f6761b5",
+  "hardware": "NVIDIA H200",
+  "tuner": "benchmark/kernels/quantization/tuning_block_wise_kernel.py",
+  "search_change": "Use BLOCK_SIZE_K=[32] for checkpoint block32 scales; upstream CUDA candidates [64,128] otherwise filter to an empty search.",
+  "tuning_slurm_job": 88009,
+  "selected_audit_and_resolver_slurm_job": 88030,
+  "validation": "39 cases against FP32 dequantized reference at rtol=0.02/atol=0.05; exact CUDA graph/eager equality; alternating timing order. BF16 rounding may differ from default tiling.",
+  "selection": "N5120/K288 batch32 restored to upstream default after timing regression.",
+  "precision": "Stock FP8 inputs and scales, BF16 output; no quantization or model changes.",
+  "splitk_tuning_slurm_job": 88079,
+  "splitk_selected_boundary_validation_slurm_job": 88084,
+  "splitk_scope": "Supported upstream SPLIT_K/SWAP_AB launch settings for N1792K5120 and N576K5120 at small verification batches; unchanged prior config selections for all M129..8192.",
+  "splitk_validation": "72/72 exploratory cases and56/56 selected boundary cases passed FP32reference rtol=.02/atol=.05 plus exact eager/graph equality; alternating timing order, selected changed cases improved1.37\u20136.22x."
+}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-h100-block32.sh b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-h100-block32.sh
index 4332962a39..b5ae175fc1 100755
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-h100-block32.sh
+++ b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-h100-block32.sh
@@ -1,5 +1,5 @@
 #!/usr/bin/env bash
 # Install the measured H100 DeepSeek-V4.1-Flash block-32 tilings into the worker's SGLang.
 set -euo pipefail
-agentic=/infmax-workspace/benchmarks/single_node/agentic
-python3 "$agentic/install_h100_block32_configs.py" "$agentic/kernel_configs/h100_dsv41_block32" /logs
+assets=/infmax-workspace/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32
+python3 "$assets/install_h100_block32_configs.py" "$assets/kernel_configs/h100_dsv41_block32" /logs
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-h200-block32.sh b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-h200-block32.sh
index 3d311f530c..6d80c7a180 100755
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-h200-block32.sh
+++ b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-h200-block32.sh
@@ -1,5 +1,5 @@
 #!/usr/bin/env bash
 # Install the measured H200 DeepSeek-V4.1-Flash block-32 tilings into the worker's SGLang.
 set -euo pipefail
-agentic=/infmax-workspace/benchmarks/single_node/agentic
-python3 "$agentic/install_h200_block32_configs.py" "$agentic/kernel_configs/h200_dsv41_block32" /logs "$DSV41_BLOCK32_TP"
+assets=/infmax-workspace/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32
+python3 "$assets/install_h200_block32_configs.py" "$assets/kernel_configs/h200_dsv41_block32" /logs "$DSV41_BLOCK32_TP"
diff --git a/benchmarks/single_node/srt-slurm-recipes/glm5.2/atom/mi355x-fp4-mtp/agentic.yaml b/benchmarks/single_node/srt-slurm-recipes/glm5.2/atom/mi355x-fp4-mtp/agentic.yaml
index e5672aa6f3..5c6f598100 100644
--- a/benchmarks/single_node/srt-slurm-recipes/glm5.2/atom/mi355x-fp4-mtp/agentic.yaml
+++ b/benchmarks/single_node/srt-slurm-recipes/glm5.2/atom/mi355x-fp4-mtp/agentic.yaml
@@ -1,6 +1,6 @@
 # GLM-5.2 MXFP4 AgentX on MI355X with ATOM MTP, TP-only with GPU-resident KV.
-# The DCP4 LMCache band stays on the legacy script: srtctl reserves ATOM's
-# kv-transfer-config for disaggregated workers.
+# The DCP4 LMCache band was removed with the legacy script (#3461): srtctl
+# reserves ATOM's kv-transfer-config for disaggregated workers.
 base:
   schema: 2
   name: glm5.2-fp4-mi355x-atom-agentic
diff --git a/benchmarks/single_node/srt-slurm-recipes/kimik3/atom/mi355x-fp4-mtp/agentic.yaml b/benchmarks/single_node/srt-slurm-recipes/kimik3/atom/mi355x-fp4-mtp/agentic.yaml
index ac5021e444..4646809ea3 100644
--- a/benchmarks/single_node/srt-slurm-recipes/kimik3/atom/mi355x-fp4-mtp/agentic.yaml
+++ b/benchmarks/single_node/srt-slurm-recipes/kimik3/atom/mi355x-fp4-mtp/agentic.yaml
@@ -1,7 +1,8 @@
 # Kimi-K3 MXFP4 AgentX on MI355X with ATOM DSpark: the interactive band, TP8
 # with GPU-resident KV and the deepest published draft (seven tokens). The
-# 1.56 TB checkpoint only fits at TP8. The DCP8 LMCache bands stay on the legacy
-# script: srtctl reserves ATOM's kv-transfer-config for disaggregated workers.
+# 1.56 TB checkpoint only fits at TP8. The DCP8 LMCache bands were removed with
+# the legacy script (#3461): srtctl reserves ATOM's kv-transfer-config for
+# disaggregated workers.
 base:
   schema: 2
   name: kimik3-fp4-mi355x-atom-agentic
diff --git a/benchmarks/single_node/srt-slurm-recipes/kimik3/vllm/mi355x-fp4-mtp/agentic.yaml b/benchmarks/single_node/srt-slurm-recipes/kimik3/vllm/mi355x-fp4-mtp/agentic.yaml
index 0621caa8e2..470956663a 100644
--- a/benchmarks/single_node/srt-slurm-recipes/kimik3/vllm/mi355x-fp4-mtp/agentic.yaml
+++ b/benchmarks/single_node/srt-slurm-recipes/kimik3/vllm/mi355x-fp4-mtp/agentic.yaml
@@ -1,8 +1,9 @@
 # Kimi-K3 MXFP4 AgentX on MI355X with vLLM DSpark
 # (https://recipes.vllm.ai/moonshotai/Kimi-K3). TP8 only: the 1.56 TB checkpoint
 # is ~195 GB per GPU. The KV cache is GPU-resident through c4 and backed by
-# vLLM's SimpleCPUOffloadConnector from c8. The DCP8 arm (c44-c70) runs without
-# a draft model and stays on the legacy script.
+# vLLM's SimpleCPUOffloadConnector from c8. The DCP8 arm (c44-c70) ran without a
+# draft model on the legacy script and was removed with it (#3461). No master
+# config uses this recipe yet.
 base:
   schema: 2
   name: kimik3-fp4-mi355x-vllm-agentic
diff --git a/benchmarks/single_node/srt-slurm-recipes/minimaxm3/atom/mi355x-fp4-mtp/agentic.yaml b/benchmarks/single_node/srt-slurm-recipes/minimaxm3/atom/mi355x-fp4-mtp/agentic.yaml
index df6b8d5165..41ad42a575 100644
--- a/benchmarks/single_node/srt-slurm-recipes/minimaxm3/atom/mi355x-fp4-mtp/agentic.yaml
+++ b/benchmarks/single_node/srt-slurm-recipes/minimaxm3/atom/mi355x-fp4-mtp/agentic.yaml
@@ -1,6 +1,7 @@
 # MiniMax-M3 MXFP4 AgentX on MI355X with ATOM EAGLE3 (GQA draft, three tokens)
-# and GPU-resident KV at TP2 and TP4. The LMCache bands stay on the legacy
-# script: srtctl reserves ATOM's kv-transfer-config for disaggregated workers.
+# and GPU-resident KV at TP2 and TP4. The LMCache bands were removed with the
+# legacy script (#3461): srtctl reserves ATOM's kv-transfer-config for
+# disaggregated workers.
 base:
   schema: 2
   name: minimaxm3-fp4-mi355x-atom-agentic
diff --git a/configs/amd-master.yaml b/configs/amd-master.yaml
index 6eeba9f6d1..e08e74208f 100644
--- a/configs/amd-master.yaml
+++ b/configs/amd-master.yaml
@@ -652,22 +652,6 @@ dsr1-fp8-mi355x-sglang-disagg-mtp:
           ep: 1
           dp-attn: false
 
-kimik3-fp4-mi355x-vllm-agentic-mtp:
-  image: vllm/vllm-openai-rocm:nightly-rocm100-7f1a5398e9610d96c473931a26c0e12bbe0d0423
-  model: moonshotai/Kimi-K3
-  model-prefix: kimik3
-  runner: cluster:mi355x-amds
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    agentic-coding:
-    - dram-utilization: 0.60
-      search-space:
-      - { tp: 8, ep: 1, dcp-size: 1, kv-offloading: none, conc-list: [1, 4] , spec-decoding: mtp }
-      - { tp: 8, ep: 1, dcp-size: 1, kv-offloading: dram, kv-offload-backend: { name: vllm-simple }, conc-list: [8, 10, 12, 14], spec-decoding: mtp }
-      - { tp: 8, ep: 1, dcp-size: 8, kv-offloading: dram, kv-offload-backend: { name: vllm-simple }, conc-list: [44, 48, 70], spec-decoding: mtp }
-
 # Kimi-K3 MXFP4 agentic-coding benchmark on MI355X via ATOM with DSpark
 # speculative decoding. Acceptance is pinned to the committed golden curve in
 # golden_al_distribution/kimik3_dspark_probabilistic_sample_method_block_rejection_sample_method.yaml
@@ -702,10 +686,6 @@ kimik3-fp4-mi355x-atom-agentic-mtp:
     - dram-utilization: 0.343
       search-space:
       - { tp: 8, kv-offloading: none, conc-list: [1, 4], spec-decoding: mtp, srt-recipe: benchmarks/single_node/srt-slurm-recipes/kimik3/atom/mi355x-fp4-mtp/agentic.yaml }
-      - { tp: 8, dcp-size: 8, kv-offloading: dram, kv-offload-backend: { name: lmcache, version: "0.4.5" }, conc-list: [14, 16, 48], spec-decoding: mtp }
-    - dram-utilization: 0.513
-      search-space:
-      - { tp: 8, dcp-size: 8, kv-offloading: dram, kv-offload-backend: { name: lmcache, version: "0.4.5" }, conc-list: [56, 72], spec-decoding: mtp }
 
 minimaxm3-fp4-mi355x-atom-agentic-mtp:
   image: rocm/atom-dev:nightly_202609171455
@@ -723,8 +703,6 @@ minimaxm3-fp4-mi355x-atom-agentic-mtp:
       search-space:
       - { tp: 4, kv-offloading: none, conc-list: [1, 2, 4, 5, 8, 10, 12, 15, 20, 24, 28, 32], spec-decoding: mtp, srt-recipe: benchmarks/single_node/srt-slurm-recipes/minimaxm3/atom/mi355x-fp4-mtp/agentic.yaml }
       - { tp: 2, kv-offloading: none, conc-list: [1, 2], spec-decoding: mtp, srt-recipe: benchmarks/single_node/srt-slurm-recipes/minimaxm3/atom/mi355x-fp4-mtp/agentic.yaml }
-      - { tp: 2, kv-offloading: dram, kv-offload-backend: { name: lmcache, version: "0.5.5rc3+rocm7.2.4" }, conc-list: [20, 25, 30], spec-decoding: mtp }
-      - { tp: 4, kv-offloading: dram, kv-offload-backend: { name: lmcache, version: "0.5.5rc3+rocm7.2.4" }, conc-list: [40, 48], spec-decoding: mtp }
 
 dsr1-fp4-mi355x-sglang-disagg:
   image: lmsysorg/sglang-rocm:v0.5.17-rocm720-mi35x-20260809
@@ -1305,21 +1283,6 @@ dsv4-fp4-mi355x-atom-disagg-agentic-lmcache-dspark:
           - "DECODE_NODES=1"
           - "DECODE_MTP_SIZE=3"
 
-minimaxm3-fp4-mi355x-vllm-agentic-mtp:
-  image: vllm/vllm-openai-rocm:nightly-2a02f6efe319c885e3ccbcecde402e0028f9ec1e
-  model: amd/MiniMax-M3-MXFP4
-  model-prefix: minimaxm3
-  runner: cluster:mi355x-amds
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    agentic-coding:
-    - dram-utilization: 0.20
-      search-space:
-      - { tp: 4, kv-offloading: none, conc-list: [1, 4, 5, 8, 10, 12, 15, 20, 24, 32], spec-decoding: mtp }
-      - { tp: 2, kv-offloading: none, conc-list: [1, 2], spec-decoding: mtp }
-
 # GLM-5.2 FP4 agentic-coding benchmark on MI355X via SGLang with MTP speculative
 # decoding. Two arms:
 # (1) TP4/EP4 with HiCache KV offloading to DRAM at concurrency [4, 8, 10, 12, 14, 16];
@@ -1392,11 +1355,9 @@ glm5.2-fp4-mi355x-atom-agentic-mtp:
     agentic-coding:
     # dram-utilization 0.171 -> TOTAL_CPU_DRAM_GB=256 per TP4 rank (recipe's
     # LMCACHE_MAX_LOCAL_CPU_SIZE=256): min(3095781,2861022)MiB*0.171*4/8 = 256.
-    # Only the DCP4 arm offloads; the other two are GPU-resident.
     - dram-utilization: 0.171
       search-space:
        - { tp: 4, kv-offloading: none, conc-list: [2, 4, 8, 10], spec-decoding: mtp, srt-recipe: benchmarks/single_node/srt-slurm-recipes/glm5.2/atom/mi355x-fp4-mtp/agentic.yaml }
-       - { tp: 4, dcp-size: 4, kv-offloading: dram, kv-offload-backend: { name: lmcache, version: "0.4.5" }, conc-list: [16, 24, 32, 40, 48], spec-decoding: mtp }
        - { tp: 8, kv-offloading: none, conc-list: [1, 2, 4], spec-decoding: mtp, srt-recipe: benchmarks/single_node/srt-slurm-recipes/glm5.2/atom/mi355x-fp4-mtp/agentic.yaml }
 
 dsv4-fp4-mi355x-sglang-agentic-mtp:
@@ -1445,22 +1406,6 @@ dsv41flash-fp4-mi355x-vllm-agentic-dspark:
       - { tp: 4, kv-offloading: none, spec-decoding: mtp, conc-list: [1, 2, 4, 8, 16, 32, 64, 128], srt-recipe: benchmarks/single_node/srt-slurm-recipes/dsv41flash/vllm/mi355x-fp4-mtp/agentic.yaml }
       - { tp: 2, kv-offloading: none, spec-decoding: mtp, conc-list: [1, 2, 4, 8, 16, 32, 64, 128], srt-recipe: benchmarks/single_node/srt-slurm-recipes/dsv41flash/vllm/mi355x-fp4-mtp/agentic.yaml }
 
-# DeepSeek-V4.1-Flash on ATOM, following the upstream TP2 / TP4 AgentX recipe.
-# Five-token DSpark uses golden AL 3.51 for throughput and real acceptance for eval.
-dsv41flash-fp4-mi355x-atom-agentic-dspark:
-  image: rocm/atom-dev:nightly_202609250902
-  model: deepseek-ai/DeepSeek-V4.1-Flash
-  model-prefix: dsv41flash
-  runner: cluster:mi355x-amds
-  precision: fp4
-  framework: atom
-  multinode: false
-  scenarios:
-    agentic-coding:
-    - search-space:
-      - { tp: 2, ep: 1, dp-attn: false, kv-offloading: none, spec-decoding: draft_model, conc-list: [1, 2, 8, 16, 32, 64] }
-      - { tp: 4, ep: 1, dp-attn: false, kv-offloading: none, spec-decoding: draft_model, conc-list: [2, 8, 16, 32, 64] }
-
 # Speculative decoding on an agentic scenario must run with simulated
 # synthetic acceptance at the committed golden AL for this model, thinking mode
 # and draft length (docs/PR_REVIEW_CHECKLIST.md), and a submission may not
diff --git a/configs/nvidia-master.yaml b/configs/nvidia-master.yaml
index 30ce89afcd..6d0ec1c375 100644
--- a/configs/nvidia-master.yaml
+++ b/configs/nvidia-master.yaml
@@ -5090,38 +5090,6 @@ qwen3.8next-fp4-b300-sglang-agentic-mtp:
     - dram-utilization: 0.8
       search-space:
       - { tp: 1, ep: 1, spec-decoding: mtp, kv-offloading: none, conc-list: [1, 4, 8, 12, 16], srt-recipe: benchmarks/single_node/srt-slurm-recipes/qwen3.8next/sglang/b300-fp4-mtp/agentic.yaml }
-# Controlled AgentX power A/B: identical software, topology, MTP settings,
-# concurrency, and memory tier across FP8 and FP4. The HBM-only rows measure
-# the natural AgentX prefix-cache workload; HiCache isolates host-tier effects.
-qwen3.5-fp8-b300-sglang-agentic-power-ab:
-  image: lmsysorg/sglang:v0.5.16-cu130
-  model: Qwen/Qwen3.5-397B-A17B-FP8
-  model-prefix: qwen3.5
-  runner: cluster:b300-nv
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    agentic-coding:
-    - dram-utilization: 0.80
-      search-space:
-      - { tp: 2, ep: 2, spec-decoding: mtp, kv-offloading: none, conc-list: [16, 32] }
-      - { tp: 2, ep: 2, spec-decoding: mtp, kv-offloading: dram, kv-offload-backend: { name: hicache }, conc-list: [16, 32] }
-
-qwen3.5-fp4-b300-sglang-agentic-power-ab:
-  image: lmsysorg/sglang:v0.5.16-cu130
-  model: nvidia/Qwen3.5-397B-A17B-NVFP4
-  model-prefix: qwen3.5
-  runner: cluster:b300-nv
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    agentic-coding:
-    - dram-utilization: 0.80
-      search-space:
-      - { tp: 2, ep: 2, spec-decoding: mtp, kv-offloading: none, conc-list: [16, 32] }
-      - { tp: 2, ep: 2, spec-decoding: mtp, kv-offloading: dram, kv-offload-backend: { name: hicache }, conc-list: [16, 32] }
 
 # CONC range conservative for H100's 80 GB HBM3 under the long-ISL with-
 # subagents corpus. hicache arm capped at conc 16 since high-conc + hicache
@@ -5291,25 +5259,6 @@ qwen3.5-fp8-h100-sglang-agentic-mtp:
       search-space:
       - { tp: 8, ep: 8, spec-decoding: mtp, kv-offloading: none, conc-list: [1, 4, 8, 12, 16], srt-recipe: benchmarks/single_node/srt-slurm-recipes/qwen3.5/sglang/h100-fp8-mtp/agentic.yaml }
       - { tp: 8, ep: 8, spec-decoding: mtp, kv-offloading: dram, kv-offload-backend: { name: hicache }, conc-list: [4, 8, 12, 16], srt-recipe: benchmarks/single_node/srt-slurm-recipes/qwen3.5/sglang/h100-fp8-mtp/agentic.yaml }
-
-# Qwen3.8-Flash-Next FP8 AgentX on H100 via SGLang with native NEXTN MTP.
-# Day-zero recipe. H100 is Hopper, so FP8: NVFP4 needs SM100 tensor cores. The
-# SGLang cookbook does not list H100, so this mirrors the H200 arm adjusted for
-# the smaller part: TP8/EP8 rather than the cookbook's TP4/EP4, since 172.8 GiB
-# at TP4 leaves too little of an 80 GB card for the 256k-capped traces.
-qwen3.8next-fp8-h100-sglang-agentic-mtp:
-  image: lmsysorg/sglang:qwen38flashnext
-  model: Qwen/Qwen3.8-Flash-Next-FP8
-  model-prefix: qwen3.8next
-  runner: cluster:h100-dgxc
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    agentic-coding:
-    - dram-utilization: 0.8
-      search-space:
-      - { tp: 8, ep: 8, spec-decoding: mtp, kv-offloading: none, conc-list: [1, 4, 8, 12, 16] }
 qwen3.5-fp4-b200-trt:
   image: nvcr.io#nvidia/tensorrt-llm/release:1.3.0rc18
   model: nvidia/Qwen3.5-397B-A17B-NVFP4
diff --git a/docs/DOCUMENTATION_PLAN.md b/docs/DOCUMENTATION_PLAN.md
index dee862d4e7..d5b91427a8 100644
--- a/docs/DOCUMENTATION_PLAN.md
+++ b/docs/DOCUMENTATION_PLAN.md
@@ -24,7 +24,7 @@ The main risks are:
 - Operational recovery knowledge remains trapped in a long agent-instruction file or historical debugging note.
 - New contributor-facing pages violate the bilingual documentation rule or omit the matching Chinese link.
 - Existing domain references are not consistently discoverable from `docs/`, and many are not paired with `_zh.md`. The migration must distinguish contributor-facing pages from internal implementation notes.
-- AgentX documentation is inconsistent: `benchmarks/single_node/agentic/README.md` describes an unpublished experimental MVP while `MODELS.md` and active master configs expose agentic-coding coverage.
+- AgentX documentation is scattered: the legacy `benchmarks/single_node/agentic/` README was removed with the bash scripts (#3461), and single-node AgentX now lives in srt-slurm recipes plus `benchmarks/srt_agentic.sh` with no maintained guide.
 
 ## Baseline knowledge inventory
 
@@ -41,7 +41,7 @@ This table records the repository state before the staged consolidation below.
 | Multi-node recipes | `benchmarks/multi_node/srt-slurm-recipes/RECIPES.md` | Recipe/master-config coupling should be visible before editing either file |
 | Runner setup | `utils/runner_setup/RUNNER_SETUP.md`, `runners/` | Provisioning and runtime launcher concerns are separated |
 | Model and hardware catalog | `MODELS.md`, `configs/*-master.yaml` | Public model list and runnable config list serve different audiences |
-| AgentX and agentic coding | `benchmarks/single_node/agentic/README.md`, `MODELS.md`, `configs/*-master.yaml` | Status and publication claims conflict. The official trace-to-result path needs one maintained guide |
+| AgentX and agentic coding | `benchmarks/single_node/srt-slurm-recipes/**/agentic.yaml`, `benchmarks/srt_agentic.sh`, `MODELS.md`, `configs/*-master.yaml` | Status and publication claims conflict. The official trace-to-result path needs one maintained guide |
 | Artifact schemas and app handoff | `utils/process_result.py`, `utils/collect_*.py`, `../InferenceX-app/.github/workflows/ingest-results.yml`, `../InferenceX-app/packages/db/src/etl/*` | Artifact identities, JSON contracts, and the boundary into InferenceX-app are not documented in this repository |
 
 ## Target information architecture
diff --git a/docs/DOCUMENTATION_PLAN_zh.md b/docs/DOCUMENTATION_PLAN_zh.md
index ee3428005f..95fa40fb59 100644
--- a/docs/DOCUMENTATION_PLAN_zh.md
+++ b/docs/DOCUMENTATION_PLAN_zh.md
@@ -24,7 +24,7 @@
 - 运维恢复知识被困在很长的 Agent 指令或历史调试记录中。
 - 新的贡献者文档违反双语规则，或没有提供对应的中文链接。
 - 现有领域参考文档没有统一从 `docs/` 发现，且许多页面没有 `_zh.md` 对应版本；迁移时必须区分贡献者文档与内部实现说明。
-- AgentX 文档存在不一致：`benchmarks/single_node/agentic/README.md` 仍描述未发布的实验性 MVP，但 `MODELS.md` 与主配置已经包含有效的 `agentic-coding` 覆盖。
+- AgentX 文档较为分散：旧的 `benchmarks/single_node/agentic/` README 已随 bash 脚本一并删除（#3461），单节点 AgentX 现由 srt-slurm 配方与 `benchmarks/srt_agentic.sh` 承载，尚无持续维护的指南。
 
 ## 基线知识清单
 
@@ -41,7 +41,7 @@
 | 多节点 Recipe | `benchmarks/multi_node/srt-slurm-recipes/RECIPES.md` | 编辑任一文件前应明确 Recipe 与主配置的耦合关系 |
 | Runner 初始化 | `utils/runner_setup/RUNNER_SETUP.md`、`runners/` | 部署与运行时启动器关注点分离 |
 | 模型与硬件目录 | `MODELS.md`、`configs/*-master.yaml` | 面向用户的模型列表与可运行配置列表服务于不同读者 |
-| AgentX 与 Agentic Coding | `benchmarks/single_node/agentic/README.md`、`MODELS.md`、`configs/*-master.yaml` | 状态与发布说明冲突，需要一份持续维护的官方 Trace 到结果指南 |
+| AgentX 与 Agentic Coding | `benchmarks/single_node/srt-slurm-recipes/**/agentic.yaml`、`benchmarks/srt_agentic.sh`、`MODELS.md`、`configs/*-master.yaml` | 状态与发布说明冲突，需要一份持续维护的官方 Trace 到结果指南 |
 | 产物 Schema 与 App 交接 | `utils/process_result.py`、`utils/collect_*.py`、`../InferenceX-app/.github/workflows/ingest-results.yml`、`../InferenceX-app/packages/db/src/etl/*` | 本仓库没有说明产物身份、JSON 契约以及进入 InferenceX-app 的边界 |
 
 ## 目标信息架构
diff --git a/docs/index.md b/docs/index.md
index a57795adc7..d0548aa27e 100644
--- a/docs/index.md
+++ b/docs/index.md
@@ -43,7 +43,7 @@ This is the mandatory low-context router for InferenceX work. Pick the one page
 | [`KLAUD_DEBUG.md`](../KLAUD_DEBUG.md) | Historical Klaud-Cold, CI, image, cluster, and GitHub CLI failure signatures |
 | [`klaud.md`](./klaud.md) / [`klaud_zh.md`](./klaud_zh.md) | Klaud Cold selection, ownership, validation and recovery |
 | [`klaud-reporting.md`](./klaud-reporting.md) / [`klaud-reporting_zh.md`](./klaud-reporting_zh.md) | Klaud PR body, progress comments, numeric comparisons and final preflight |
-| [`benchmarks/single_node/agentic/README.md`](../benchmarks/single_node/agentic/README.md) | AgentX trace benchmark implementation |
+| [`benchmarks/srt_agentic.sh`](../benchmarks/srt_agentic.sh) | AgentX trace replay client shared by single- and multi-node srt-slurm recipes |
 
 ## Context rules
 
diff --git a/docs/index_zh.md b/docs/index_zh.md
index 5c1b3da8bb..d8c6026163 100644
--- a/docs/index_zh.md
+++ b/docs/index_zh.md
@@ -43,7 +43,7 @@
 | [`KLAUD_DEBUG.md`](../KLAUD_DEBUG.md) | Klaud-Cold、CI、镜像、集群与 GitHub CLI 的历史故障特征 |
 | [`klaud.md`](./klaud.md) / [`klaud_zh.md`](./klaud_zh.md) | Klaud Cold 候选选择、所有权、验证与恢复 |
 | [`klaud-reporting.md`](./klaud-reporting.md) / [`klaud-reporting_zh.md`](./klaud-reporting_zh.md) | Klaud PR 正文、进度评论、数值对比与最终预检 |
-| [`benchmarks/single_node/agentic/README.md`](../benchmarks/single_node/agentic/README.md) | AgentX Trace 基准测试实现 |
+| [`benchmarks/srt_agentic.sh`](../benchmarks/srt_agentic.sh) | 单节点与多节点 srt-slurm 配方共用的 AgentX Trace 回放客户端 |
 
 ## 上下文规则
 
diff --git a/docs/waiver/2508.md b/docs/waiver/2508.md
index 6fdc3e7c64..66e9e9afa4 100644
--- a/docs/waiver/2508.md
+++ b/docs/waiver/2508.md
@@ -9,8 +9,9 @@ prohibits unless covered by a filled-out waiver.
 - **Master config entry:** `kimik3-fp4-mi355x-vllm-agentic-mtp` in [`configs/amd-master.yaml`](../../configs/amd-master.yaml)
 - **Pinned image:** `vllm/vllm-openai-rocm:nightly-cb8104839c141609d99f1254459ef3a4f1bd4263`
 - **Patch entrypoint:** `bash "$(dirname "$0")/apply_k3_container_patches.sh"` invoked from
-  [`benchmarks/single_node/agentic/kimik3_fp4_mi355x_mtp.sh`](../../benchmarks/single_node/agentic/kimik3_fp4_mi355x_mtp.sh)
-- **Patch script:** [`benchmarks/single_node/agentic/apply_k3_container_patches.sh`](../../benchmarks/single_node/agentic/apply_k3_container_patches.sh)
+  the legacy `kimik3_fp4_mi355x_mtp.sh` AgentX script
+- **Patch script:** `apply_k3_container_patches.sh`
+- **Status:** both scripts, and the `kimik3-fp4-mi355x-vllm-agentic-mtp` config that used them, were removed with the legacy single-node AgentX bash path (#3460, #3461).
 
 ## What is patched
 
diff --git a/runners/launch_gb200-nv.sh b/runners/launch_gb200-nv.sh
index 66a0996ba3..c217d42f0a 100755
--- a/runners/launch_gb200-nv.sh
+++ b/runners/launch_gb200-nv.sh
@@ -106,41 +106,6 @@ if [[ "$IS_MULTINODE" != true && -n "${SRT_RECIPE:-}" ]]; then
     exit $?
 fi
 
-# Direct single-tray AgentX uses the existing shared image and HF caches.
-if [[ "$MODEL_PREFIX" == "dsv41flash" && ( "$FRAMEWORK" == "vllm" || "$FRAMEWORK" == "sglang" ) && "${IS_MULTINODE}" != "true" ]]; then
-    check_env_vars SPEC_DECODING
-    BENCH_SCRIPT="benchmarks/single_node/agentic/${MODEL_PREFIX}_${PRECISION}_gb200_${FRAMEWORK}"
-    case "$SPEC_DECODING" in
-        mtp) BENCH_SCRIPT+="_mtp.sh" ;;
-        none)
-            [[ "$FRAMEWORK" == "sglang" ]] || { echo "Native STP requires the SGLang recipe" >&2; exit 1; }
-            BENCH_SCRIPT+=".sh"
-            ;;
-        *) echo "Unsupported SPEC_DECODING=$SPEC_DECODING" >&2; exit 1 ;;
-    esac
-    # Cover DSpark5 verification for concurrent AgentX subagents at c1/c2/c4.
-    export DSV41_MIN_CUDAGRAPH_CAPTURE_SIZE=64
-    [[ "${IS_AGENTIC}" == "1" && -f "$BENCH_SCRIPT" ]] || {
-        echo "Unsupported single-node recipe: $BENCH_SCRIPT" >&2
-        exit 1
-    }
-    HF_HUB_CACHE_HOST_PATH="/mnt/lustre01/users-public/sa-shared/hf-hub-cache"
-    mkdir -p "$HF_HUB_CACHE_HOST_PATH"
-    export MODEL_PATH="$MODEL" HF_HUB_CACHE=/hf-cache
-    export INFMAX_CONTAINER_WORKSPACE=/ix RESULT_DIR=/ix/results
-    SQUASH_FILE="$SQUASH_DIR/$(echo "$IMAGE" | sed 's/[\/:@#]/_/g').sqsh"
-    import_squash "$SQUASH_FILE" "$IMAGE"
-    srun --account="$SLURM_ACCOUNT" --partition="$SLURM_PARTITION" \
-        --nodes=1 --ntasks=1 --gpus="${TP:?}" --exclusive --mem=0 \
-        --time="${SALLOC_TIME_LIMIT}" --job-name="$RUNNER_NAME" \
-        --mpi=none --container-image="$SQUASH_FILE" \
-        --container-mounts="$GITHUB_WORKSPACE:/ix,$HF_HUB_CACHE_HOST_PATH:/hf-cache" \
-        --no-container-mount-home --container-remap-root \
-        --container-workdir=/ix --no-container-entrypoint \
-        --export=ALL,PORT=8888 bash "$BENCH_SCRIPT"
-    exit $?
-fi
-
 if [[ "$FRAMEWORK" == "llmd-vllm" ]]; then
     if [[ "$MODEL_PREFIX" == "dsv4" && "$PRECISION" == "fp4" ]]; then
         export MODEL_PATH="/mnt/numa1/models/DeepSeek-V4-Pro"
diff --git a/runners/launch_gb300-nv.sh b/runners/launch_gb300-nv.sh
index 7d579742b0..a09a238e40 100644
--- a/runners/launch_gb300-nv.sh
+++ b/runners/launch_gb300-nv.sh
@@ -117,39 +117,6 @@ if [[ "$IS_MULTINODE" != true && -n "${SRT_RECIPE:-}" ]]; then
         --var DYNAMO_WHEELS_CACHE_HOST_PATH "$DYNAMO_WHEELS_CACHE_HOST_PATH"
     exit $?
 fi
-# Keep this branch before the nginx import and srtctl setup.
-if [[ "$MODEL_PREFIX" == "dsv41flash" && ( "$FRAMEWORK" == "vllm" || "$FRAMEWORK" == "sglang" ) && "${IS_MULTINODE}" != "true" ]]; then
-    check_env_vars SPEC_DECODING
-    BENCH_SCRIPT="benchmarks/single_node/agentic/${MODEL_PREFIX}_${PRECISION}_gb300_${FRAMEWORK}"
-    case "$SPEC_DECODING" in
-        mtp) BENCH_SCRIPT+="_mtp.sh" ;;
-        none)
-            [[ "$FRAMEWORK" == "sglang" ]] || { echo "Native STP requires the SGLang recipe" >&2; exit 1; }
-            BENCH_SCRIPT+=".sh"
-            ;;
-        *) echo "Unsupported SPEC_DECODING=$SPEC_DECODING" >&2; exit 1 ;;
-    esac
-    # Cover DSpark5 verification for concurrent AgentX subagents at c1/c2/c4.
-    export DSV41_MIN_CUDAGRAPH_CAPTURE_SIZE=64
-    [[ "${IS_AGENTIC}" == "1" && -f "$BENCH_SCRIPT" ]] || {
-        echo "Unsupported single-node recipe: $BENCH_SCRIPT" >&2
-        exit 1
-    }
-    export HF_HUB_CACHE=/hf-cache
-    export INFMAX_CONTAINER_WORKSPACE=/ix
-    export RESULT_DIR=/ix/results
-    # Cold model loading and graph capture exceeded the one-hour frontend deadline.
-    export VLLM_ENGINE_READY_TIMEOUT_S=7200
-    srun --account="$SLURM_ACCOUNT" --partition="$SLURM_PARTITION" \
-        --nodes=1 --ntasks=1 --gpus="${TP:?}" --cpus-per-task=144 --exclusive --mem=0 \
-        --time="${SALLOC_TIME_LIMIT}" --job-name="$RUNNER_NAME" \
-        --mpi=none --container-image="$SQUASH_FILE" \
-        --container-mounts="$GITHUB_WORKSPACE:/ix,$HF_HUB_CACHE_HOST_PATH:/hf-cache" \
-        --no-container-mount-home --container-remap-root \
-        --container-workdir=/ix --no-container-entrypoint \
-        --export=ALL,PORT=8888 bash "$BENCH_SCRIPT"
-    exit $?
-fi
 
 import_squash "$NGINX_SQUASH_FILE" "$NGINX_IMAGE"
 
diff --git a/runners/launch_mi355x-amds.sh b/runners/launch_mi355x-amds.sh
index 29282f9cc8..7f313f8fb9 100644
--- a/runners/launch_mi355x-amds.sh
+++ b/runners/launch_mi355x-amds.sh
@@ -369,13 +369,6 @@ else
         BENCHMARK_SCRIPT="$SCRIPT_FALLBACK"
     fi
 
-    if [[ "$BENCHMARK_SCRIPT" == "benchmarks/single_node/agentic/minimaxm3_fp4_mi355x_atom_mtp.sh" ]]; then
-        export MODEL_PATH="$MODEL"
-        export ENABLE_PREFIX_CACHING=true
-        export AITER_LOG_LEVEL=WARNING
-        export EVAL_TASKS_DIR=infx/evals/gsm8k.yaml
-    fi
-
     srun --jobid=$JOB_ID \
         --container-image=$SQUASH_FILE \
         --container-mounts=$GITHUB_WORKSPACE:$CONTAINER_REPO/,$HF_HUB_CACHE_MOUNT:$HF_HUB_CACHE,$AIPERF_MMAP_CACHE_HOST_PATH:/aiperf_mmap_cache \

From f45d6bff02f51c43ae9c686817ce58d0b483ed42 Mon Sep 17 00:00:00 2001
From: functionstackx <47992694+functionstackx@users.noreply.github.com>
Date: Sat, 26 Sep 2026 02:29:41 -0400
Subject: [PATCH 2/6] Drop the DSv4.1 Flash H100/H200 block32 kernel tilings
 instead of restoring them

Per review: do not re-add the block32 installers and kernel configs.
Remove the restored dsv41flash-block32 assets, the two block32 setup scripts,
the recipes' setup_script entries and the H200 DSV41_BLOCK32_TP env vars; the
H100/H200 SGLang AgentX recipes now run with SGLang's default tilings.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
---
 .../install_h100_block32_configs.py           |  48 -------
 .../install_h200_block32_configs.py           |  43 ------
 ...3,dtype=fp8_w8a8,block_shape=[32, 32].json | 114 ----------------
 ...3,dtype=fp8_w8a8,block_shape=[32, 32].json |  50 -------
 ...3,dtype=fp8_w8a8,block_shape=[32, 32].json |  98 --------------
 ...3,dtype=fp8_w8a8,block_shape=[32, 32].json |  66 ----------
 ...3,dtype=fp8_w8a8,block_shape=[32, 32].json |  66 ----------
 ...3,dtype=fp8_w8a8,block_shape=[32, 32].json | 122 ------------------
 .../h100_dsv41_block32/provenance.json        |  71 ----------
 ...0,dtype=fp8_w8a8,block_shape=[32, 32].json |  58 ---------
 ...0,dtype=fp8_w8a8,block_shape=[32, 32].json |  66 ----------
 ...0,dtype=fp8_w8a8,block_shape=[32, 32].json |  66 ----------
 ...0,dtype=fp8_w8a8,block_shape=[32, 32].json | 122 ------------------
 .../h200_dsv41_block32/provenance.json        |  16 ---
 .../configs/dsv41flash-h100-block32.sh        |   5 -
 .../configs/dsv41flash-h200-block32.sh        |   5 -
 .../sglang/h100-fp4-mtp/agentic.yaml          |   2 -
 .../sglang/h200-fp4-mtp/agentic.yaml          |  35 -----
 18 files changed, 1053 deletions(-)
 delete mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/install_h100_block32_configs.py
 delete mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/install_h200_block32_configs.py
 delete mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=1280,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
 delete mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=1792,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
 delete mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=512,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
 delete mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=5120,K=1024,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
 delete mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=5120,K=288,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
 delete mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=576,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
 delete mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/provenance.json
 delete mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=1792,K=5120,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json
 delete mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=5120,K=1024,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json
 delete mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=5120,K=288,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json
 delete mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=576,K=5120,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json
 delete mode 100644 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/provenance.json
 delete mode 100755 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-h100-block32.sh
 delete mode 100755 benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-h200-block32.sh

diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/install_h100_block32_configs.py b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/install_h100_block32_configs.py
deleted file mode 100644
index 88c3fc97f6..0000000000
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/install_h100_block32_configs.py
+++ /dev/null
@@ -1,48 +0,0 @@
-"""Install the H100 V4.1 tiling configs into the imported SGLang kernel package."""  # noqa: INP001 -- Standalone recipe executable.
-
-import hashlib
-import json
-import shutil
-import sys
-from pathlib import Path
-
-from sglang.kernels.ops.quantization import fp8_kernel
-
-
-def main() -> None:
-    source = Path(sys.argv[1])
-    artifacts = Path(sys.argv[2]) / "fp8_kernel_configs"
-    device = fp8_kernel.get_device_name().replace(" ", "_")
-    if device != "NVIDIA_H100_80GB_HBM3":
-        raise RuntimeError(f"H100 tiling configs cannot be installed on {device}")
-    destination = Path(fp8_kernel.__file__).resolve().parent / "configs"
-    destination.mkdir(exist_ok=True)
-    artifacts.mkdir(parents=True, exist_ok=True)
-    for n, k in [
-        (5120, 1024),
-        (5120, 288),
-        (576, 5120),
-        (1792, 5120),
-        (512, 5120),
-        (1280, 5120),
-    ]:
-        name = (
-            f"N={n},K={k},device_name={device},dtype=fp8_w8a8,block_shape=[32, 32].json"
-        )
-        path = source / name
-        expected = {
-            int(m): config for m, config in json.loads(path.read_text()).items()
-        }
-        shutil.copyfile(path, destination / name)
-        shutil.copyfile(path, artifacts / name)
-        fp8_kernel.get_w8a8_block_fp8_configs.cache_clear()
-        actual = fp8_kernel.get_w8a8_block_fp8_configs(n, k, 32, 32)
-        if actual != expected:
-            raise RuntimeError(f"SGLang did not resolve the installed config: {name}")
-        print(
-            f"Installed {destination / name}: sha256={hashlib.sha256(path.read_bytes()).hexdigest()}"
-        )
-
-
-if __name__ == "__main__":
-    main()
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/install_h200_block32_configs.py b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/install_h200_block32_configs.py
deleted file mode 100644
index 3d34a865e2..0000000000
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/install_h200_block32_configs.py
+++ /dev/null
@@ -1,43 +0,0 @@
-"""Install the H200 V4.1 tiling configs into the imported SGLang kernel package."""  # noqa: INP001 -- Standalone recipe executable.
-
-import hashlib
-import json
-import shutil
-import sys
-from pathlib import Path
-
-from sglang.kernels.ops.quantization import fp8_kernel
-
-
-def main() -> None:
-    source = Path(sys.argv[1])
-    artifacts = Path(sys.argv[2]) / "fp8_kernel_configs"
-    tp = int(sys.argv[3])
-    if tp not in (4, 8):
-        raise ValueError(f"H200 tiling configs support TP4 or TP8, got TP{tp}")
-    device = fp8_kernel.get_device_name().replace(" ", "_")
-    if device != "NVIDIA_H200":
-        raise RuntimeError(f"H200 tiling configs cannot be installed on {device}")
-    destination = Path(fp8_kernel.__file__).resolve().parent / "configs"
-    destination.mkdir(exist_ok=True)
-    artifacts.mkdir(parents=True, exist_ok=True)
-    shapes = [(1792, 5120)]
-    if tp == 8:
-        shapes += [(5120, 1024), (5120, 288), (576, 5120)]
-    for n, k in shapes:
-        name = f"N={n},K={k},device_name={device},dtype=fp8_w8a8,block_shape=[32, 32].json"
-        path = source / name
-        expected = {int(m): config for m, config in json.loads(path.read_text()).items()}
-        shutil.copyfile(path, destination / name)
-        shutil.copyfile(path, artifacts / name)
-        fp8_kernel.get_w8a8_block_fp8_configs.cache_clear()
-        actual = fp8_kernel.get_w8a8_block_fp8_configs(n, k, 32, 32)
-        if actual != expected:
-            raise RuntimeError(f"SGLang did not resolve the installed config: {name}")
-        print(
-            f"Installed {destination / name}: sha256={hashlib.sha256(path.read_bytes()).hexdigest()}"
-        )
-
-
-if __name__ == "__main__":
-    main()
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=1280,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=1280,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
deleted file mode 100644
index 971110754f..0000000000
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=1280,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json	
+++ /dev/null
@@ -1,114 +0,0 @@
-{
-  "0": {
-    "BLOCK_SIZE_M": 64,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3
-  },
-  "1": {
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3,
-    "SPLIT_K": 16,
-    "SWAP_AB": false
-  },
-  "5": {
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3,
-    "SPLIT_K": 16,
-    "SWAP_AB": false
-  },
-  "10": {
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3,
-    "SPLIT_K": 16,
-    "SWAP_AB": true
-  },
-  "20": {
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3,
-    "SPLIT_K": 8,
-    "SWAP_AB": true
-  },
-  "40": {
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3,
-    "SPLIT_K": 8,
-    "SWAP_AB": true
-  },
-  "80": {
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3,
-    "SPLIT_K": 16,
-    "SWAP_AB": true
-  },
-  "160": {
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3,
-    "SPLIT_K": 8,
-    "SWAP_AB": true
-  },
-  "190": {
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3,
-    "SPLIT_K": 8,
-    "SWAP_AB": true
-  },
-  "191": {
-    "BLOCK_SIZE_M": 64,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3
-  },
-  "320": {
-    "BLOCK_SIZE_M": 64,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3
-  },
-  "321": {
-    "BLOCK_SIZE_M": 64,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3
-  }
-}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=1792,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=1792,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
deleted file mode 100644
index 986ff1953b..0000000000
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=1792,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json	
+++ /dev/null
@@ -1,50 +0,0 @@
-{
-  "1": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 16,
-    "SWAP_AB": false,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "6": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 16,
-    "SWAP_AB": true,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "24": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 64,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 8,
-    "SWAP_AB": true,
-    "num_stages": 4,
-    "num_warps": 4
-  },
-  "96": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 64,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 8,
-    "SWAP_AB": true,
-    "num_stages": 4,
-    "num_warps": 4
-  },
-  "128": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 64,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "num_stages": 3,
-    "num_warps": 4
-  }
-}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=512,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=512,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
deleted file mode 100644
index 3c4f7326b0..0000000000
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=512,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json	
+++ /dev/null
@@ -1,98 +0,0 @@
-{
-  "0": {
-    "BLOCK_SIZE_M": 64,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3
-  },
-  "1": {
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3,
-    "SPLIT_K": 16,
-    "SWAP_AB": false
-  },
-  "5": {
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3,
-    "SPLIT_K": 16,
-    "SWAP_AB": false
-  },
-  "10": {
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3,
-    "SPLIT_K": 16,
-    "SWAP_AB": false
-  },
-  "20": {
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3,
-    "SPLIT_K": 16,
-    "SWAP_AB": true
-  },
-  "40": {
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3,
-    "SPLIT_K": 16,
-    "SWAP_AB": true
-  },
-  "80": {
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3,
-    "SPLIT_K": 8,
-    "SWAP_AB": true
-  },
-  "160": {
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3,
-    "SPLIT_K": 4,
-    "SWAP_AB": true
-  },
-  "320": {
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3,
-    "SPLIT_K": 4,
-    "SWAP_AB": true
-  },
-  "321": {
-    "BLOCK_SIZE_M": 64,
-    "BLOCK_SIZE_N": 32,
-    "BLOCK_SIZE_K": 32,
-    "GROUP_SIZE_M": 32,
-    "num_warps": 4,
-    "num_stages": 3
-  }
-}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=5120,K=1024,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=5120,K=1024,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
deleted file mode 100644
index fece264886..0000000000
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=5120,K=1024,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json	
+++ /dev/null
@@ -1,66 +0,0 @@
-{
-    "1": {
-        "BLOCK_SIZE_M": 16,
-        "BLOCK_SIZE_N": 64,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 16,
-        "num_warps": 4,
-        "num_stages": 5
-    },
-    "4": {
-        "BLOCK_SIZE_M": 16,
-        "BLOCK_SIZE_N": 64,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 64,
-        "num_warps": 4,
-        "num_stages": 4
-    },
-    "16": {
-        "BLOCK_SIZE_M": 16,
-        "BLOCK_SIZE_N": 32,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 1,
-        "num_warps": 4,
-        "num_stages": 4
-    },
-    "32": {
-        "BLOCK_SIZE_M": 16,
-        "BLOCK_SIZE_N": 64,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 16,
-        "num_warps": 4,
-        "num_stages": 5
-    },
-    "64": {
-        "BLOCK_SIZE_M": 64,
-        "BLOCK_SIZE_N": 128,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 1,
-        "num_warps": 4,
-        "num_stages": 5
-    },
-    "256": {
-        "BLOCK_SIZE_M": 64,
-        "BLOCK_SIZE_N": 128,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 16,
-        "num_warps": 4,
-        "num_stages": 5
-    },
-    "1024": {
-        "BLOCK_SIZE_M": 64,
-        "BLOCK_SIZE_N": 128,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 1,
-        "num_warps": 4,
-        "num_stages": 5
-    },
-    "4096": {
-        "BLOCK_SIZE_M": 64,
-        "BLOCK_SIZE_N": 128,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 1,
-        "num_warps": 4,
-        "num_stages": 4
-    }
-}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=5120,K=288,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=5120,K=288,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
deleted file mode 100644
index ccf51c6e0a..0000000000
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=5120,K=288,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json	
+++ /dev/null
@@ -1,66 +0,0 @@
-{
-    "1": {
-        "BLOCK_SIZE_M": 16,
-        "BLOCK_SIZE_N": 64,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 16,
-        "num_warps": 4,
-        "num_stages": 2
-    },
-    "4": {
-        "BLOCK_SIZE_M": 16,
-        "BLOCK_SIZE_N": 128,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 32,
-        "num_warps": 4,
-        "num_stages": 2
-    },
-    "16": {
-        "BLOCK_SIZE_M": 16,
-        "BLOCK_SIZE_N": 64,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 32,
-        "num_warps": 8,
-        "num_stages": 2
-    },
-    "32": {
-        "BLOCK_SIZE_M": 32,
-        "BLOCK_SIZE_N": 32,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 64,
-        "num_warps": 4,
-        "num_stages": 5
-    },
-    "64": {
-        "BLOCK_SIZE_M": 64,
-        "BLOCK_SIZE_N": 32,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 32,
-        "num_warps": 4,
-        "num_stages": 3
-    },
-    "256": {
-        "BLOCK_SIZE_M": 64,
-        "BLOCK_SIZE_N": 128,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 64,
-        "num_warps": 4,
-        "num_stages": 5
-    },
-    "1024": {
-        "BLOCK_SIZE_M": 64,
-        "BLOCK_SIZE_N": 128,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 16,
-        "num_warps": 4,
-        "num_stages": 2
-    },
-    "4096": {
-        "BLOCK_SIZE_M": 64,
-        "BLOCK_SIZE_N": 128,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 16,
-        "num_warps": 4,
-        "num_stages": 3
-    }
-}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=576,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=576,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json
deleted file mode 100644
index 4e45c66e6d..0000000000
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/N=576,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json	
+++ /dev/null
@@ -1,122 +0,0 @@
-{
-  "1": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 16,
-    "SWAP_AB": false,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "4": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 16,
-    "SWAP_AB": true,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "6": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 16,
-    "SWAP_AB": true,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "16": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 16,
-    "SWAP_AB": true,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "24": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 16,
-    "SWAP_AB": true,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "32": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 16,
-    "SWAP_AB": true,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "64": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 64,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 8,
-    "SWAP_AB": true,
-    "num_stages": 4,
-    "num_warps": 4
-  },
-  "96": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 64,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 8,
-    "SWAP_AB": true,
-    "num_stages": 4,
-    "num_warps": 4
-  },
-  "128": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 64,
-    "GROUP_SIZE_M": 16,
-    "num_stages": 5,
-    "num_warps": 4
-  },
-  "193": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 64,
-    "GROUP_SIZE_M": 32,
-    "num_stages": 4,
-    "num_warps": 4
-  },
-  "256": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 64,
-    "GROUP_SIZE_M": 32,
-    "num_stages": 4,
-    "num_warps": 4
-  },
-  "1024": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 64,
-    "BLOCK_SIZE_N": 128,
-    "GROUP_SIZE_M": 1,
-    "num_stages": 4,
-    "num_warps": 4
-  },
-  "4096": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 64,
-    "BLOCK_SIZE_N": 128,
-    "GROUP_SIZE_M": 1,
-    "num_stages": 5,
-    "num_warps": 4
-  }
-}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/provenance.json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/provenance.json
deleted file mode 100644
index 9c4fff096c..0000000000
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h100_dsv41_block32/provenance.json
+++ /dev/null
@@ -1,71 +0,0 @@
-{
-  "sglang_commit": "0f6761b54facebb47f2068f87ecccd8f14da3a0e",
-  "image": "lmsysorg/sglang:nightly-dev-cu13-20260921-0f6761b5",
-  "hardware": "NVIDIA H100 80GB HBM3",
-  "tuner": "benchmark/kernels/quantization/tuning_block_wise_kernel.py",
-  "search_change": "Use BLOCK_SIZE_K=[32] for checkpoint block32 scales; upstream CUDA candidates [64,128] otherwise filter to an empty search.",
-  "tuning_slurm_job": 18452,
-  "selected_audit_slurm_job": 18457,
-  "resolver_install_slurm_job": 18458,
-  "validation": "39 cases: FP32 dequantized reference rtol=0.02/atol=0.05; exact CUDA graph/eager equality; alternating timing order. BF16 rounding may differ from default tiling.",
-  "selection": "N5120/K288 batch64 restored to upstream default after timing regression.",
-  "precision": "Stock FP8 inputs and scales, BF16 output; no quantization or model changes.",
-  "split_k_audit": {
-    "slurm_job": 18479,
-    "cases": 80,
-    "mechanism": "Pinned upstream Hopper SPLIT_K and SWAP_AB launch configuration, FP32 partial reduction, unchanged FP8 inputs/scales and BF16 output",
-    "shapes": [
-      "N1792/K5120",
-      "N576/K5120"
-    ],
-    "boundary_validation": "Slurm18480:56/56 FP32-reference and exact CUDA-graph replay checks passed, batch boundariesM1..1024; changed cases all faster in alternating-order microtimings",
-    "large_batch_policy": "Retain original fallback selection for M129..8192 using explicit boundary anchors"
-  },
-  "draft_projection_audit": {
-    "device": "NVIDIA H100 80GB HBM3",
-    "image": "nightly-dev-cu13-20260922-582389ce",
-    "search": {
-      "job": 18492,
-      "cases": 144,
-      "status": "COMPLETED0:0"
-    },
-    "boundary": {
-      "job": 18493,
-      "cases": 206,
-      "status": "COMPLETED0:0"
-    },
-    "trimmed_boundary": {
-      "job": 18494,
-      "cases": 20,
-      "status": "COMPLETED0:0"
-    },
-    "reference": "FP32 dequant reference rtol.02/atol.05; exact eager/graph replay; no quantization or kernel code changes",
-    "ranges": {
-      "N512": "M1..320 tuned; M0 and M>=321 stock",
-      "N1280": "M1..190 tuned; M0 and M>=191 stock"
-    },
-    "retained_speedup_range": [
-      1.1861693666195017,
-      6.770688638312714
-    ],
-    "baseline_l2_max": 0.0016977186314761639,
-    "candidate_l2_max": 0.0016970767173916101,
-    "hashes": {
-      "N=1280,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json": "635f684b6ed8b538709249baa5c474bf8583f952bcee330e01fd0f9e0ce714a3",
-      "N=512,K=5120,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[32, 32].json": "193fc012550dce39bb4de0a4bde28eed5cee79f2900ed95841fe18ccb6d8c187"
-    },
-    "serving_qualification": {
-      "candidate_run": 35683668032,
-      "baseline_run": 35683669703,
-      "protocol": "Matched canonical C1, 3600 seconds, normal warmup; zero profiled errors and valid power in both runs",
-      "candidate_total_tps_per_gpu": 1239.57314,
-      "baseline_total_tps_per_gpu": 958.51774,
-      "candidate_p90_interactivity": 90.89516,
-      "baseline_p90_interactivity": 60.93631,
-      "full_gsm8k_run": 35683671358,
-      "full_gsm8k_examples": 1319,
-      "full_gsm8k_strict_accuracy": 0.9696739954510993,
-      "final_sweep": "Required after integration"
-    }
-  }
-}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=1792,K=5120,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=1792,K=5120,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json
deleted file mode 100644
index 83620a6b10..0000000000
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=1792,K=5120,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json	
+++ /dev/null
@@ -1,58 +0,0 @@
-{
-  "1": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 64,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 8,
-    "SWAP_AB": true,
-    "num_stages": 4,
-    "num_warps": 4
-  },
-  "2": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 64,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "6": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 16,
-    "SWAP_AB": true,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "24": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 8,
-    "SWAP_AB": true,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "96": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 16,
-    "SWAP_AB": true,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "128": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 64,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "num_stages": 3,
-    "num_warps": 4
-  }
-}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=5120,K=1024,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=5120,K=1024,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json
deleted file mode 100644
index 850f61947e..0000000000
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=5120,K=1024,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json	
+++ /dev/null
@@ -1,66 +0,0 @@
-{
-    "1": {
-        "BLOCK_SIZE_M": 16,
-        "BLOCK_SIZE_N": 64,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 64,
-        "num_warps": 4,
-        "num_stages": 3
-    },
-    "4": {
-        "BLOCK_SIZE_M": 16,
-        "BLOCK_SIZE_N": 32,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 64,
-        "num_warps": 4,
-        "num_stages": 4
-    },
-    "16": {
-        "BLOCK_SIZE_M": 16,
-        "BLOCK_SIZE_N": 64,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 16,
-        "num_warps": 4,
-        "num_stages": 5
-    },
-    "32": {
-        "BLOCK_SIZE_M": 32,
-        "BLOCK_SIZE_N": 64,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 32,
-        "num_warps": 4,
-        "num_stages": 4
-    },
-    "64": {
-        "BLOCK_SIZE_M": 32,
-        "BLOCK_SIZE_N": 128,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 16,
-        "num_warps": 4,
-        "num_stages": 3
-    },
-    "256": {
-        "BLOCK_SIZE_M": 64,
-        "BLOCK_SIZE_N": 128,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 16,
-        "num_warps": 4,
-        "num_stages": 5
-    },
-    "1024": {
-        "BLOCK_SIZE_M": 64,
-        "BLOCK_SIZE_N": 128,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 32,
-        "num_warps": 4,
-        "num_stages": 3
-    },
-    "4096": {
-        "BLOCK_SIZE_M": 64,
-        "BLOCK_SIZE_N": 128,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 1,
-        "num_warps": 4,
-        "num_stages": 4
-    }
-}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=5120,K=288,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=5120,K=288,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json
deleted file mode 100644
index 94b8a3fccf..0000000000
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=5120,K=288,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json	
+++ /dev/null
@@ -1,66 +0,0 @@
-{
-    "1": {
-        "BLOCK_SIZE_M": 32,
-        "BLOCK_SIZE_N": 64,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 32,
-        "num_warps": 8,
-        "num_stages": 2
-    },
-    "4": {
-        "BLOCK_SIZE_M": 64,
-        "BLOCK_SIZE_N": 128,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 1,
-        "num_warps": 4,
-        "num_stages": 4
-    },
-    "16": {
-        "BLOCK_SIZE_M": 32,
-        "BLOCK_SIZE_N": 128,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 64,
-        "num_warps": 4,
-        "num_stages": 2
-    },
-    "32": {
-        "BLOCK_SIZE_M": 64,
-        "BLOCK_SIZE_N": 32,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 32,
-        "num_warps": 4,
-        "num_stages": 3
-    },
-    "64": {
-        "BLOCK_SIZE_M": 16,
-        "BLOCK_SIZE_N": 128,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 16,
-        "num_warps": 4,
-        "num_stages": 4
-    },
-    "256": {
-        "BLOCK_SIZE_M": 64,
-        "BLOCK_SIZE_N": 128,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 64,
-        "num_warps": 4,
-        "num_stages": 3
-    },
-    "1024": {
-        "BLOCK_SIZE_M": 64,
-        "BLOCK_SIZE_N": 128,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 32,
-        "num_warps": 4,
-        "num_stages": 4
-    },
-    "4096": {
-        "BLOCK_SIZE_M": 64,
-        "BLOCK_SIZE_N": 128,
-        "BLOCK_SIZE_K": 32,
-        "GROUP_SIZE_M": 64,
-        "num_warps": 4,
-        "num_stages": 4
-    }
-}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=576,K=5120,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=576,K=5120,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json
deleted file mode 100644
index 4359ec7f93..0000000000
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/N=576,K=5120,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[32, 32].json	
+++ /dev/null
@@ -1,122 +0,0 @@
-{
-  "1": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 16,
-    "SWAP_AB": false,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "4": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 16,
-    "SWAP_AB": false,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "6": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 16,
-    "SWAP_AB": false,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "16": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 16,
-    "SWAP_AB": true,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "24": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 16,
-    "SWAP_AB": true,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "32": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 16,
-    "SWAP_AB": true,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "64": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 8,
-    "SWAP_AB": true,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "96": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 32,
-    "GROUP_SIZE_M": 32,
-    "SPLIT_K": 8,
-    "SWAP_AB": true,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "128": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 64,
-    "GROUP_SIZE_M": 16,
-    "num_stages": 5,
-    "num_warps": 4
-  },
-  "193": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 64,
-    "GROUP_SIZE_M": 64,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "256": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 16,
-    "BLOCK_SIZE_N": 64,
-    "GROUP_SIZE_M": 64,
-    "num_stages": 3,
-    "num_warps": 4
-  },
-  "1024": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 64,
-    "BLOCK_SIZE_N": 128,
-    "GROUP_SIZE_M": 1,
-    "num_stages": 5,
-    "num_warps": 4
-  },
-  "4096": {
-    "BLOCK_SIZE_K": 32,
-    "BLOCK_SIZE_M": 64,
-    "BLOCK_SIZE_N": 128,
-    "GROUP_SIZE_M": 1,
-    "num_stages": 5,
-    "num_warps": 4
-  }
-}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/provenance.json b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/provenance.json
deleted file mode 100644
index 074329f645..0000000000
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32/kernel_configs/h200_dsv41_block32/provenance.json
+++ /dev/null
@@ -1,16 +0,0 @@
-{
-  "sglang_commit": "0f6761b54facebb47f2068f87ecccd8f14da3a0e",
-  "image": "lmsysorg/sglang:nightly-dev-cu13-20260921-0f6761b5",
-  "hardware": "NVIDIA H200",
-  "tuner": "benchmark/kernels/quantization/tuning_block_wise_kernel.py",
-  "search_change": "Use BLOCK_SIZE_K=[32] for checkpoint block32 scales; upstream CUDA candidates [64,128] otherwise filter to an empty search.",
-  "tuning_slurm_job": 88009,
-  "selected_audit_and_resolver_slurm_job": 88030,
-  "validation": "39 cases against FP32 dequantized reference at rtol=0.02/atol=0.05; exact CUDA graph/eager equality; alternating timing order. BF16 rounding may differ from default tiling.",
-  "selection": "N5120/K288 batch32 restored to upstream default after timing regression.",
-  "precision": "Stock FP8 inputs and scales, BF16 output; no quantization or model changes.",
-  "splitk_tuning_slurm_job": 88079,
-  "splitk_selected_boundary_validation_slurm_job": 88084,
-  "splitk_scope": "Supported upstream SPLIT_K/SWAP_AB launch settings for N1792K5120 and N576K5120 at small verification batches; unchanged prior config selections for all M129..8192.",
-  "splitk_validation": "72/72 exploratory cases and56/56 selected boundary cases passed FP32reference rtol=.02/atol=.05 plus exact eager/graph equality; alternating timing order, selected changed cases improved1.37\u20136.22x."
-}
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-h100-block32.sh b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-h100-block32.sh
deleted file mode 100755
index b5ae175fc1..0000000000
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-h100-block32.sh
+++ /dev/null
@@ -1,5 +0,0 @@
-#!/usr/bin/env bash
-# Install the measured H100 DeepSeek-V4.1-Flash block-32 tilings into the worker's SGLang.
-set -euo pipefail
-assets=/infmax-workspace/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32
-python3 "$assets/install_h100_block32_configs.py" "$assets/kernel_configs/h100_dsv41_block32" /logs
diff --git a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-h200-block32.sh b/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-h200-block32.sh
deleted file mode 100755
index 6d80c7a180..0000000000
--- a/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-h200-block32.sh
+++ /dev/null
@@ -1,5 +0,0 @@
-#!/usr/bin/env bash
-# Install the measured H200 DeepSeek-V4.1-Flash block-32 tilings into the worker's SGLang.
-set -euo pipefail
-assets=/infmax-workspace/benchmarks/multi_node/srt-slurm-recipes/configs/dsv41flash-block32
-python3 "$assets/install_h200_block32_configs.py" "$assets/kernel_configs/h200_dsv41_block32" /logs "$DSV41_BLOCK32_TP"
diff --git a/benchmarks/single_node/srt-slurm-recipes/dsv41flash/sglang/h100-fp4-mtp/agentic.yaml b/benchmarks/single_node/srt-slurm-recipes/dsv41flash/sglang/h100-fp4-mtp/agentic.yaml
index 6f4767f0ce..041fb4849f 100644
--- a/benchmarks/single_node/srt-slurm-recipes/dsv41flash/sglang/h100-fp4-mtp/agentic.yaml
+++ b/benchmarks/single_node/srt-slurm-recipes/dsv41flash/sglang/h100-fp4-mtp/agentic.yaml
@@ -29,8 +29,6 @@ base:
   health_check:
     interval_seconds: 10
     max_attempts: 360
-  # Hardware-specific tiling only; checkpoint data, scales and dtypes are unchanged.
-  setup_script: dsv41flash-h100-block32.sh
   roles:
     agg:
       nodes: 1
diff --git a/benchmarks/single_node/srt-slurm-recipes/dsv41flash/sglang/h200-fp4-mtp/agentic.yaml b/benchmarks/single_node/srt-slurm-recipes/dsv41flash/sglang/h200-fp4-mtp/agentic.yaml
index 0b55113278..a5d86251de 100644
--- a/benchmarks/single_node/srt-slurm-recipes/dsv41flash/sglang/h200-fp4-mtp/agentic.yaml
+++ b/benchmarks/single_node/srt-slurm-recipes/dsv41flash/sglang/h200-fp4-mtp/agentic.yaml
@@ -19,9 +19,6 @@ base:
     tachometer:
       enabled: false
   engine: sglang
-  # Measured H200 launch configurations: the replicated projection on TP4 and
-  # TP8, plus the qualified TP8 sharded shapes.
-  setup_script: dsv41flash-h200-block32.sh
   roles:
     agg:
       nodes: 1
@@ -90,8 +87,6 @@ override_tp4_c1:
       args:
         tensor-parallel-size: 4
         max-running-requests: 2
-      env:
-        DSV41_BLOCK32_TP: '4'
   benchmark:
     env:
       CONC: '1'
@@ -104,8 +99,6 @@ override_tp4_c2:
       args:
         tensor-parallel-size: 4
         max-running-requests: 4
-      env:
-        DSV41_BLOCK32_TP: '4'
   benchmark:
     env:
       CONC: '2'
@@ -118,8 +111,6 @@ override_tp4_c4:
       args:
         tensor-parallel-size: 4
         max-running-requests: 8
-      env:
-        DSV41_BLOCK32_TP: '4'
   benchmark:
     env:
       CONC: '4'
@@ -132,8 +123,6 @@ override_tp4_c8:
       args:
         tensor-parallel-size: 4
         max-running-requests: 16
-      env:
-        DSV41_BLOCK32_TP: '4'
   benchmark:
     env:
       CONC: '8'
@@ -146,8 +135,6 @@ override_tp4_c16:
       args:
         tensor-parallel-size: 4
         max-running-requests: 32
-      env:
-        DSV41_BLOCK32_TP: '4'
   benchmark:
     env:
       CONC: '16'
@@ -160,8 +147,6 @@ override_tp4_c32:
       args:
         tensor-parallel-size: 4
         max-running-requests: 64
-      env:
-        DSV41_BLOCK32_TP: '4'
   benchmark:
     env:
       CONC: '32'
@@ -174,8 +159,6 @@ override_tp4_c64:
       args:
         tensor-parallel-size: 4
         max-running-requests: 64
-      env:
-        DSV41_BLOCK32_TP: '4'
   benchmark:
     env:
       CONC: '64'
@@ -188,8 +171,6 @@ override_tp4_c128:
       args:
         tensor-parallel-size: 4
         max-running-requests: 64
-      env:
-        DSV41_BLOCK32_TP: '4'
   benchmark:
     env:
       CONC: '128'
@@ -202,8 +183,6 @@ override_tp8_c1:
       args:
         tensor-parallel-size: 8
         max-running-requests: 2
-      env:
-        DSV41_BLOCK32_TP: '8'
   benchmark:
     env:
       CONC: '1'
@@ -216,8 +195,6 @@ override_tp8_c2:
       args:
         tensor-parallel-size: 8
         max-running-requests: 4
-      env:
-        DSV41_BLOCK32_TP: '8'
   benchmark:
     env:
       CONC: '2'
@@ -230,8 +207,6 @@ override_tp8_c4:
       args:
         tensor-parallel-size: 8
         max-running-requests: 8
-      env:
-        DSV41_BLOCK32_TP: '8'
   benchmark:
     env:
       CONC: '4'
@@ -244,8 +219,6 @@ override_tp8_c8:
       args:
         tensor-parallel-size: 8
         max-running-requests: 16
-      env:
-        DSV41_BLOCK32_TP: '8'
   benchmark:
     env:
       CONC: '8'
@@ -258,8 +231,6 @@ override_tp8_c16:
       args:
         tensor-parallel-size: 8
         max-running-requests: 32
-      env:
-        DSV41_BLOCK32_TP: '8'
   benchmark:
     env:
       CONC: '16'
@@ -272,8 +243,6 @@ override_tp8_c32:
       args:
         tensor-parallel-size: 8
         max-running-requests: 64
-      env:
-        DSV41_BLOCK32_TP: '8'
   benchmark:
     env:
       CONC: '32'
@@ -286,8 +255,6 @@ override_tp8_c64:
       args:
         tensor-parallel-size: 8
         max-running-requests: 64
-      env:
-        DSV41_BLOCK32_TP: '8'
   benchmark:
     env:
       CONC: '64'
@@ -300,8 +267,6 @@ override_tp8_c128:
       args:
         tensor-parallel-size: 8
         max-running-requests: 64
-      env:
-        DSV41_BLOCK32_TP: '8'
   benchmark:
     env:
       CONC: '128'

From 3fdd26d2d64b42ca597fc70990e5a054effe1cfa Mon Sep 17 00:00:00 2001
From: functionstackx <47992694+functionstackx@users.noreply.github.com>
Date: Sat, 26 Sep 2026 02:35:28 -0400
Subject: [PATCH 3/6] Drop the KLAUD_DEBUG script pointer (fixed_seq_len is
 deleted in #3464)

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
---
 KLAUD_DEBUG.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/KLAUD_DEBUG.md b/KLAUD_DEBUG.md
index 0f0fb21043..7d50415bc7 100644
--- a/KLAUD_DEBUG.md
+++ b/KLAUD_DEBUG.md
@@ -77,7 +77,7 @@ changing the benchmark recipe.
 
 **Fix:** in `benchmarks/single_node/<recipe>.sh`, either:
 1. **Lower `--gpu-memory-utilization`** (`0.95 → 0.90`, sometimes 0.85). Matches the H100/H200/B200 NVIDIA pattern. Smallest blast radius.
-2. **Disable the profiler entirely** for cases where lowering isn't enough: `export VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS=0` before `vllm serve`. Matches `benchmarks/single_node/fixed_seq_len/deprecated/kimik2.5_fp4_b200.sh`.
+2. **Disable the profiler entirely** for cases where lowering isn't enough: `export VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS=0` before `vllm serve`.
 
 Seen on: #1395 (kimik2.5-fp4-b200-vllm, needed env var), #1403 (gptoss-fp4-mi300x-vllm, needed 0.90), #1461 (dsv4-fp8-h200-vllm, needed 0.90).
 

From 43cc8af7b5e0e21bd7d7d25e8baaab307022215b Mon Sep 17 00:00:00 2001
From: functionstackx <47992694+functionstackx@users.noreply.github.com>
Date: Sat, 26 Sep 2026 02:34:42 -0400
Subject: [PATCH 4/6] Delete configs/deprecated and drop remaining
 fixed_seq_len pointers

- Delete configs/deprecated/{amd,nvidia}-master.yaml; nothing reads them.
  AGENTS.md, docs/configuration-procedures and the Claude review prompt now
  say retired entries are deleted, with git history and perf-changelog.yaml
  as the record.
- Rewrite the add-model-hardware playbook and configuration procedures for
  srt-slurm recipes instead of bash scripts under fixed_seq_len/.
- Point MODELS links, the KLAUD_DEBUG example and a deprecated multi-node
  comment away from the deleted folders.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
---
 .claude/commands/add-model-hardware.md        |    76 +-
 .github/workflows/claude.yml                  |     2 +-
 AGENTS.md                                     |     6 +-
 MODELS.md                                     |    12 +-
 MODELS_zh.md                                  |    12 +-
 .../amd_utils/deprecated/models_vllm.yaml     |     4 +-
 configs/deprecated/amd-master.yaml            |  3464 ----
 configs/deprecated/nvidia-master.yaml         | 16588 ----------------
 docs/configuration-procedures.md              |    10 +-
 docs/configuration-procedures_zh.md           |    10 +-
 10 files changed, 61 insertions(+), 20123 deletions(-)
 delete mode 100644 configs/deprecated/amd-master.yaml
 delete mode 100644 configs/deprecated/nvidia-master.yaml

diff --git a/.claude/commands/add-model-hardware.md b/.claude/commands/add-model-hardware.md
index 0100361404..02c1efc043 100644
--- a/.claude/commands/add-model-hardware.md
+++ b/.claude/commands/add-model-hardware.md
@@ -1,5 +1,5 @@
 ---
-description: Add a new model+hardware single-node benchmark recipe (script + master-config entry + perf-changelog + launcher routing), open a [Klaud Cold] PR, label full-sweep-fail-fast, and monitor CI
+description: Add a new model+hardware single-node benchmark recipe (srt-slurm recipe + master-config entry + perf-changelog), open a [Klaud Cold] PR, label full-sweep-fail-fast, and monitor CI
 argument-hint: <model-link> <gpu-sku> [recipes-link] [draft-model-link] [mtp]
 ---
 
@@ -40,45 +40,43 @@ Check `MODELS.md` before choosing a model, scenario, or precision. Do not reintr
 
 **A. In-codebase research (primary because this repo is the source of truth):**
 ```bash
-# similar benchmark scripts: same model on other SKUs, AND same SKU on other models
-ls benchmarks/single_node/fixed_seq_len/<model>_*.sh benchmarks/single_node/fixed_seq_len/*_<sku>*.sh
+# similar srt-slurm recipes: same model on other SKUs, AND same SKU on other models
+ls -d benchmarks/single_node/srt-slurm-recipes/<model>/*/* benchmarks/single_node/srt-slurm-recipes/*/*/<sku>-*
 # similar master-config entries (search spaces, image, parallelism), this model + analogues
 grep -nE "<model>-|.*-<sku>-" configs/{nvidia,amd}-master.yaml
-# the runner launcher for this SKU (script-name routing, env, mounts, MODEL_PATH rewrite)
-sed -n '1,80p' runners/launch_<sku>*.sh
-# shared helpers the scripts rely on
-grep -nE "run_benchmark_serving|setup_eval_context|wait_for_server_ready|start_gpu_monitor" benchmarks/benchmark_lib.sh
+# how a matrix point selects exactly one recipe variant (TP, GPUs, CONC, KV_OFFLOADING, image)
+sed -n '/def select_recipe/,/^def runtime_arguments/p' infx/srt_slurm/single_node.py
 ```
-- **Read multiple sibling scripts** end-to-end for the exact env vars and serve shape (`VLLM_*`,
+- **Read multiple sibling recipes** end-to-end for the exact engine args, env vars and serve shape (`VLLM_*`,
   `SGLANG_*`, device mapping, download/cache handling, `--enforce-eager` vs graph capture,
-  KV-cache dtype, attention/MoE backend, parsers). These are the truth for each runner.
+  KV-cache dtype, attention/MoE backend, parsers, `setup_script`). These are the truth for each runner.
 - **Compare several master-config search spaces** (e.g. `dsr1`, `qwen3.5`, the same model on a
   sibling SKU) to choose `{tp, ep, dp-attn} × concurrency` combos that fit *this* hardware's
   memory. Small-memory SKUs like h100/mi300x go TP8-only, while bigger SKUs add tp4/tp2/DEP.
 - **Internalize the fixed-seq-len nuances from the existing configs**: `8k1k` runs do **not**
   need the full `MAX_MODEL_LEN` (the matrix supplies `isl + osl + slack`), and graph-capture
   batch sizes are scaled to concurrency/scenario (and spec-token count for MTP), not maxed.
-  Copy how sibling scripts/configs already do it.
+  Copy how sibling recipes/configs already do it.
 
 **B. External research (confirm against upstream guidance):**
 - **`WebFetch` the model-link card + its `config.json`** → confirm `model:` id, precision, max
   context, architecture, spec-decode fields (`num_mtp_modules`, etc.).
 - **`WebFetch` the recipes-link** (if given) → canonical `vllm serve` flags + troubleshooting.
-  Reconcile with what the sibling scripts do. If they conflict, follow the repo and note why.
+  Reconcile with what the sibling recipes do. If they conflict, follow the repo and note why.
 - If a **draft-model-link** is given, note its id for `--speculative-config` and check the card
   for method (`eagle3` vs native `mtp`) and recommended token count.
 - Pick the **image tag** from the sibling's master-config entry (or recipes page) and **verify
   it exists** on the registry before using it.
 
-This research directly feeds Step 2 (script flags/env) and Step 3 (search space).
+This research directly feeds Step 2 (recipe args/env) and Step 3 (search space).
 
-## What you're producing (4–5 files)
+## What you're producing (3 files)
 
-1. `benchmarks/single_node/fixed_seq_len/<model>_<precision>_<sku>[_<engine>][_mtp].sh`
+1. `benchmarks/single_node/srt-slurm-recipes/<model-prefix>/<engine>/<sku>-<precision>[-mtp]/8k1k.yaml`
+   (an `agentic.yaml` beside it for AgentX)
 2. an entry in either master config, **`configs/nvidia-master.yaml`** (b*/h*/gb* SKUs) or
-   **`configs/amd-master.yaml`** (mi* SKUs)
+   **`configs/amd-master.yaml`** (mi* SKUs), with `srt-recipe:` on every search-space row
 3. a `perf-changelog.yaml` entry (this diff vs main is what selects the sweep)
-4. (if missing) `SPEC_SUFFIX`/framework-suffix routing in `runners/launch_<sku>*.sh`
 
 ## Step 1 — branch + find the sibling to copy
 
@@ -86,17 +84,20 @@ This research directly feeds Step 2 (script flags/env) and Step 3 (search space)
 git checkout main && git pull origin main
 git checkout -b feat/<model>-<sku>[-mtp]-dayzero
 # nearest sibling: same model other SKU, or same SKU other model
-ls benchmarks/single_node/fixed_seq_len/<model>_*           # same model, other hardware
-ls benchmarks/single_node/fixed_seq_len/*_<sku>*.sh         # same hardware, other model
+ls -d benchmarks/single_node/srt-slurm-recipes/<model>/*/*      # same model, other hardware
+ls -d benchmarks/single_node/srt-slurm-recipes/*/*/<sku>-*       # same hardware, other model
 grep -n "<model>-<precision>-<sku>" configs/{nvidia,amd}-master.yaml
 ```
-Read the closest sibling script **and** its master-config entry. Copy their flag shapes and
+Read the closest sibling recipe **and** its master-config entry. Copy their flag shapes and
 search-space structure rather than inventing. The right model is "same model on a sibling SKU,
 adjusted for this hardware's quirks."
 
-## Step 2 — write the benchmark script
+## Step 2 — write the srt-slurm recipe
 
-Copy the sibling script and adjust. Things that vary and must be checked against the sibling /
+Copy the sibling recipe (`base:` plus one `override_*` variant per matrix point) and adjust.
+Engine flags are `roles.agg.args` keys without the leading `--`; env vars go in
+`roles.agg.env`; each variant names its `CONC` (and `KV_OFFLOADING` for AgentX) in
+`benchmark.env`, and `model.container` must equal the master-config `image`. Things that vary and must be checked against the sibling /
 the model's `recipes.vllm.ai` page:
 - **Mandatory model flags** (carry from the sibling): block size, parser flags
   (`--tool-call-parser` / `--reasoning-parser`), `--language-model-only` for text-only sweeps,
@@ -113,7 +114,7 @@ the model's `recipes.vllm.ai` page:
 - **Memory headroom.** Bigger checkpoints constrain TP/EP. If the sibling on a smaller-memory
   SKU is TP8-only (e.g. h100), match that.
 
-Validate as you go: `bash -n <script>`.
+Validate as you go: `python3 -c "import yaml; yaml.safe_load(open('<recipe>'))"`.
 
 ## Step 3 — master-config entry + search space
 
@@ -127,18 +128,12 @@ Append `<model>-<precision>-<sku>[-<engine>][-mtp]` after the sibling, with the
 
 Confirm which master file by SKU: `mi*` → `amd-master.yaml`, everything else → `nvidia-master.yaml`.
 
-## Step 4 — launcher routing
+## Step 4 — no launcher routing
 
-The runner's launcher must resolve your script name. Most build it as
-`<model>_<precision>_<sku>[_<framework>][_mtp].sh`. h200 launchers already carry the framework
-+ `SPEC_SUFFIX`. **h100 and mi300x/mi355x launchers have historically hardcoded the bare
-`_<sku>.sh`**. Check and fix this if you added a framework-tagged or `_mtp` script:
-```bash
-grep -n 'SPEC_SUFFIX\|FRAMEWORK_SUFFIX\|bash benchmarks\|EXP_NAME%%' runners/launch_<sku>*.sh
-```
-If needed, add `SPEC_SUFFIX=$([[ "$SPEC_DECODING" == "mtp" ]] && printf '_mtp' || printf '')`
-(and/or the framework suffix) near the top and splice it into the bench-script path. Simulate
-both `none`/`mtp` to confirm the resolved filename exists.
+Single-node points with an `srt-recipe:` go through `launch_srt_single_node`, which picks the
+one recipe variant whose TP/GPU count, `CONC`, `KV_OFFLOADING` and image match the matrix
+point (`infx/srt_slurm/single_node.py::select_recipe`). No per-script launcher routing is
+needed; if a point matches zero or several variants, fix the recipe, not the launcher.
 
 ## Step 5 — perf-changelog
 
@@ -149,7 +144,7 @@ new entry is **required** for CI to run your config.
 ## Step 6 — validate locally
 
 ```bash
-bash -n benchmarks/single_node/fixed_seq_len/<script>
+python3 -c "import yaml; yaml.safe_load(open('benchmarks/single_node/srt-slurm-recipes/<recipe>'))"
 python3 -c "import yaml; yaml.safe_load(open('configs/<nvidia|amd>-master.yaml')); yaml.safe_load(open('perf-changelog.yaml'))"
 uv run --no-project --exclude-newer PT12H --python 3.12 --with pydantic --with pyyaml \
   utils/matrix_logic/generate_sweep_configs.py test-config \
@@ -186,14 +181,11 @@ external **`Inferact/MiniMax-M3-EAGLE3`** draft, `method: eagle3`, **3 speculati
 ```
 - **CUDA (b*/h*).** Pin the drafter to `FLASH_ATTN` because FlashInfer can't run the MHA EAGLE3 head
   at the mandatory page-size 128. Scale cudagraph capture to `CONC * (1 + NUM_SPEC_TOKENS)`.
-- **ROCm (mi*)**: no backend pin (server runs `TRITON_ATTN`). The shipped
-  `vllm/vllm-openai-rocm:minimax-m3` image's AMD model lacks `SupportsEagle3`, so until the
-  upstream fix (`vllm-project/vllm#45546`) is in the image, patch the installed
-  `models/minimax_m3/amd/model.py` in-place before serving. Copy the idempotent, drift-checked
-  `python3 - <<'PYEOF' ... PYEOF` block verbatim from `minimaxm3_fp8_mi355x_mtp.sh` (adds
-  `EagleModelMixin` + aux-hidden-state emission + `SupportsEagle3` on the two outer classes).
-- **All**: route benchmark prompts through `--use-chat-template` (+ `pip install -q datasets
-  pandas`). Raw random tokens tank spec-decode acceptance. Search space mirrors the non-MTP
+- **ROCm (mi*)**: no backend pin (server runs `TRITON_ATTN`). Use an image that already carries
+  the upstream `SupportsEagle3` fix (`vllm-project/vllm#45546`); the old in-place model patch was
+  retired with the legacy bash scripts. Copy the sibling `minimaxm3/vllm/mi*-mtp` recipe.
+- **All**: set `USE_CHAT_TEMPLATE: 'true'` in the recipe's `benchmark.env` (the single-node
+  adapter requires it whenever speculation is on). Raw random tokens tank spec-decode acceptance. Search space mirrors the non-MTP
   entry trimmed at the extreme-conc end, latency rows starting at conc 1, `tp2-ep2` dropped.
 - Other models may instead use **native MTP** (`method: mtp`, no external draft) when the
   checkpoint ships MTP modules (`num_mtp_modules > 0`), e.g. the DeepSeek-V4 recipes.
diff --git a/.github/workflows/claude.yml b/.github/workflows/claude.yml
index c14aa3c440..72cc766ae1 100644
--- a/.github/workflows/claude.yml
+++ b/.github/workflows/claude.yml
@@ -611,5 +611,5 @@ jobs:
             **Deprecation validation:**
             - Reject new active entries for retired models, scenarios, or precisions unless `MODELS.md` explicitly documents an exception.
             - Preserve the GLM-5.1 B200 TileRT exception and the conditional non-speculative AgentX policy; do not infer retirement from the existence of a speculative counterpart.
-            - Deprecated entries belong only in `configs/deprecated/amd-master.yaml` or `configs/deprecated/nvidia-master.yaml`, following `AGENTS.md`. Historical entries in these archives are not active submissions.
+            - Retired entries are deleted from the master configs, not archived, following `AGENTS.md`; git history and `perf-changelog.yaml` are the historical record.
             - For a newly supported model, require matching updates to `MODELS.md` and `MODELS_zh.md` rather than rejecting it against an obsolete model list.
diff --git a/AGENTS.md b/AGENTS.md
index 20d2291b81..44c887e843 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -52,12 +52,10 @@ check_env_vars IS_MULTINODE MODEL_NAME PRECISION
 
 ## Deprecating benchmark configs
 
-- Move deprecated entries out of the active master config into [`configs/deprecated/amd-master.yaml`](configs/deprecated/amd-master.yaml) or [`configs/deprecated/nvidia-master.yaml`](configs/deprecated/nvidia-master.yaml), matching the vendor. These are the only deprecated master-config files; do not create separate files per model, scenario, or deprecation.
-- For a partial deprecation, archive only the retired scenarios and retain the supported scenarios in the active entry. Preserve archived settings and explanatory comments; do not update historical image pins or runners during archival.
-- Keep every archive key unique. If a key already exists with different settings, preserve both versions with a descriptive suffix on the historical key and a comment recording its original config key. Existing colliding 1k1k versions use `-deprecated-1k1k`. Never overwrite an archived version or add duplicate YAML keys.
+- Delete retired entries from the active master config; do not archive them. Git history and `perf-changelog.yaml` are the record of past settings. For a partial deprecation, remove only the retired scenarios and retain the supported scenarios in the active entry.
 - Check retirement statements in [`MODELS.md`](MODELS.md) against active configs and script routing in the same PR, and update `MODELS.md` plus `MODELS_zh.md` together. Preserve explicitly documented exceptions and conditional retirement policies; do not treat planned retirement as completed.
 - Remove unused retired-model branches from launchers and runtime settings, and update workflow/agent guidance that still recommends retired coverage. Audit callers before removing shared helpers; retained SPEED-Bench collectors and historical result readers may still need model-specific support.
-- Keep these archives out of active sweep inputs. Follow the existing benchmark-script archival convention, moving retired scripts into the sibling `deprecated/` directory only when no active config still uses them.
+- Delete recipes, setup scripts and other assets that no active config uses any more rather than moving them to a `deprecated/` directory.
 
 ## Runner launchers (one file per pool)
 
diff --git a/MODELS.md b/MODELS.md
index ce8e6f83f0..9e6fe11d53 100644
--- a/MODELS.md
+++ b/MODELS.md
@@ -12,7 +12,7 @@ InferenceX-e2e runs on a fixed, limited pool of GPUs and is maintained by a smal
 
 **Monday, August 3, 2026** is the last day for the scenarios and precisions in the first table below. They are deprecated after that date. The separate A/B baseline retirement is described below that table.
 
-**Partially enacted on 2026-08-04** in [#2493](https://github.com/SemiAnalysisAI/InferenceX/pull/2493): the scenario and precision retirements in the first table were carried out. This removed 54 config keys from the active master configs and archived them under [`configs/deprecated/`](configs/deprecated/), with their benchmark scripts moved to the sibling `deprecated/` directories. The speculative-decoding A/B retirements in the second table are **not yet enacted**. See the note under that table.
+**Partially enacted on 2026-08-04** in [#2493](https://github.com/SemiAnalysisAI/InferenceX/pull/2493): the scenario and precision retirements in the first table were carried out. This removed 54 config keys from the active master configs and archived them under `configs/deprecated/`, with their benchmark scripts moved to the sibling `deprecated/` directories. The speculative-decoding A/B retirements in the second table are **not yet enacted**. See the note under that table.
 
 Scenario and precision retirements:
 
@@ -42,7 +42,7 @@ The A/B retirements below concern standalone non-spec-decode baselines maintaine
 
 **Thursday, August 6, 2026** is the last day for the **Single-turn 8k1k** scenario on **Kimi-K2.5/2.6/2.7-Code** (`kimik2.5`). The scenario is deprecated for these models after that date. Rationale: Kimi-K3 launched on July 27, 2026, so GPU cluster time shifts to the newer frontier model. Combined with the Agentic coding deprecation above, this leaves `kimik2.5` with no active scenario. The model is **fully retired after August 6, 2026**.
 
-**Enacted on 2026-08-07** in [#2527](https://github.com/SemiAnalysisAI/InferenceX/pull/2527): 17 `kimik2.5` config keys were removed from the active master configs and archived under [`configs/deprecated/`](configs/deprecated/), now consolidated in `configs/deprecated/nvidia-master.yaml` (10) and `configs/deprecated/amd-master.yaml` (7), and their 12 benchmark scripts were moved to the sibling `deprecated/` directories. `kimik2.5` now has **no active configuration in any master config** and is fully retired. The same PR archived `kimik2.5-int4-h100-vllm`, an agentic-coding key that #2493 left behind in `nvidia-master.yaml` after moving its script to `benchmarks/single_node/agentic/deprecated/`. It is now in `configs/deprecated/nvidia-master.yaml` with its siblings. The SPEED-Bench acceptance-length script `benchmarks/single_node/speedbench/kimik2.5_fp4_b300_vllm.sh` is intentionally kept. Speedbench is driven by `speedbench-al.yml`, not the master configs, matching how #2493 treated MiniMax-M3.
+**Enacted on 2026-08-07** in [#2527](https://github.com/SemiAnalysisAI/InferenceX/pull/2527): 17 `kimik2.5` config keys were removed from the active master configs and archived under `configs/deprecated/`, now consolidated in `configs/deprecated/nvidia-master.yaml` (10) and `configs/deprecated/amd-master.yaml` (7), and their 12 benchmark scripts were moved to the sibling `deprecated/` directories. `kimik2.5` now has **no active configuration in any master config** and is fully retired. The same PR archived `kimik2.5-int4-h100-vllm`, an agentic-coding key that #2493 left behind in `nvidia-master.yaml` after moving its script to `benchmarks/single_node/agentic/deprecated/`. It is now in `configs/deprecated/nvidia-master.yaml` with its siblings. The SPEED-Bench acceptance-length script `benchmarks/single_node/speedbench/kimik2.5_fp4_b300_vllm.sh` is intentionally kept. Speedbench is driven by `speedbench-al.yml`, not the master configs, matching how #2493 treated MiniMax-M3.
 
 ### Tuesday, September 8, 2026
 
@@ -54,9 +54,9 @@ The A/B retirements below concern standalone non-spec-decode baselines maintaine
 
 Rationale: `dsv4` carries the largest single-turn footprint in the repository. 45 active config keys use the 8k1k scenario, 32 in `configs/nvidia-master.yaml` and 13 in `configs/amd-master.yaml`, spanning H200, B200, B300, GB200, GB300, MI300X, MI325X, and MI355X across vLLM, SGLang, TensorRT-LLM, ATOM, Dynamo, and llm-d. That is a large share of every full sweep. AgentX trace replay is the scenario AI labs and the ML community ask about, and DeepSeek-V4-Pro's 19 agentic config keys are the part of `dsv4` that feeds the published North Star Pareto frontier. Retiring the fixed-sequence-length arm frees cluster hours for AgentX and for new frontier models such as Qwen3.8-Flash-Next without reducing what we publish for this model. Single-turn 8k1k stays active for the other models that still list it.
 
-**Enacted on 2026-09-09** in [#2921](https://github.com/SemiAnalysisAI/InferenceX/pull/2921): 46 `dsv4` 8k1k config keys were removed from the active master configs and archived under [`configs/deprecated/`](configs/deprecated/), now consolidated in `configs/deprecated/nvidia-master.yaml` (33) and `configs/deprecated/amd-master.yaml` (13), and their 28 benchmark scripts were moved to the sibling `deprecated/` directories (25 under `benchmarks/single_node/fixed_seq_len/`, 3 under `benchmarks/multi_node/`), matching how [#2493](https://github.com/SemiAnalysisAI/InferenceX/pull/2493) and [#2527](https://github.com/SemiAnalysisAI/InferenceX/pull/2527) were carried out. The count is 46 rather than the 45 quoted above because `dsv4-fp4-b200-dynamo-sglang` landed after this notice was written. The 19 agentic-coding keys are untouched: `dsv4` continues to run and publish with agentic coding as its only scenario. The SPEED-Bench acceptance-length scripts for `dsv4` are intentionally kept. Speedbench is driven by `speedbench-al.yml`, not the master configs. The srt-slurm and llm-d recipe YAMLs referenced by the archived multi-node keys stay in place as inert reference data, as #2493 and #2527 left theirs.
+**Enacted on 2026-09-09** in [#2921](https://github.com/SemiAnalysisAI/InferenceX/pull/2921): 46 `dsv4` 8k1k config keys were removed from the active master configs and archived under `configs/deprecated/`, now consolidated in `configs/deprecated/nvidia-master.yaml` (33) and `configs/deprecated/amd-master.yaml` (13), and their 28 benchmark scripts were moved to the sibling `deprecated/` directories (25 under `benchmarks/single_node/fixed_seq_len/`, 3 under `benchmarks/multi_node/`), matching how [#2493](https://github.com/SemiAnalysisAI/InferenceX/pull/2493) and [#2527](https://github.com/SemiAnalysisAI/InferenceX/pull/2527) were carried out. The count is 46 rather than the 45 quoted above because `dsv4-fp4-b200-dynamo-sglang` landed after this notice was written. The 19 agentic-coding keys are untouched: `dsv4` continues to run and publish with agentic coding as its only scenario. The SPEED-Bench acceptance-length scripts for `dsv4` are intentionally kept. Speedbench is driven by `speedbench-al.yml`, not the master configs. The srt-slurm and llm-d recipe YAMLs referenced by the archived multi-node keys stay in place as inert reference data, as #2493 and #2527 left theirs.
 
-**Deprecation parity audit (2026-09-21):** Active master configs and benchmark-script locations match the enacted retirements above and in the support matrix below. GLM-5.1 B200 TileRT remains the documented exception to the earlier GLM-5/5.1 and 1k1k retirements. Conditional A/B baseline retirement remains pending; non-speculative Pareto contributors remain supported. The broader routing audit also removed stale retired-model branches from launchers/runtime settings and a GLM-5-only environment override, and corrected workflow/agent guidance that still recommended retired coverage. SPEED-Bench collectors, historical result readers, and the explicitly retained recipe YAMLs remain available. Deprecated configs are consolidated in [`configs/deprecated/amd-master.yaml`](configs/deprecated/amd-master.yaml) and [`configs/deprecated/nvidia-master.yaml`](configs/deprecated/nvidia-master.yaml).
+**Deprecation parity audit (2026-09-21):** Active master configs and benchmark-script locations match the enacted retirements above and in the support matrix below. GLM-5.1 B200 TileRT remains the documented exception to the earlier GLM-5/5.1 and 1k1k retirements. Conditional A/B baseline retirement remains pending; non-speculative Pareto contributors remain supported. The broader routing audit also removed stale retired-model branches from launchers/runtime settings and a GLM-5-only environment override, and corrected workflow/agent guidance that still recommended retired coverage. SPEED-Bench collectors, historical result readers, and the explicitly retained recipe YAMLs remain available. Deprecated configs were consolidated in `configs/deprecated/amd-master.yaml` and `configs/deprecated/nvidia-master.yaml`.
 
 **Single-node SRT-only cutover (2026-09-22):** Active single-node fixed-sequence recipes now use SRT-Slurm. The two Docker-only Qwen3.5 RTX PRO 6000 FP4 configs (with and without MTP) are retired, with their original settings preserved in `configs/deprecated/nvidia-master.yaml` and their scripts in `benchmarks/single_node/fixed_seq_len/deprecated/`. The unused `rtx6000pro-lat` runner mappings, launcher, and runtime settings are removed. Qwen3.5 remains active on the other supported Slurm pools; AgentX and multi-node coverage are unchanged.
 
@@ -66,7 +66,7 @@ Rationale: `dsv4` carries the largest single-turn footprint in the repository. 4
 |---|---|---|
 | Agentic coding | Long Context, Multi Turn Realistic traffic trace replay with sub agents | Active. This is the trace-replay agentic-coding benchmark (see the srt-slurm recipes under [`benchmarks/single_node/srt-slurm-recipes/`](benchmarks/single_node/srt-slurm-recipes/) and the shared client [`benchmarks/srt_agentic.sh`](benchmarks/srt_agentic.sh)). Going forward, new models will likely be onboarded with agentic coding only. Speculative decoding may be enabled or disabled to produce the best Pareto points; a separate non-spec-decode A/B baseline is not required (see [Deprecation Notice](#deprecation-notice)). |
 | Single-turn 8k1k | 8192 / 1024 | Active. This is the primary fixed-sequence-length scenario. |
-| Single-turn 1k1k | 1024 / 1024 | Deprecated since 2026-07-17 ([#2263](https://github.com/SemiAnalysisAI/InferenceX/pull/2263)), to save GPU cluster time for higher-priority real-world agentic-coding benchmarks and new frontier models. Archived configs live in [`configs/deprecated/`](configs/deprecated/). The GLM-5.1 B200 TileRT point added later in [#2533](https://github.com/SemiAnalysisAI/InferenceX/pull/2533) remains active. |
+| Single-turn 1k1k | 1024 / 1024 | Deprecated since 2026-07-17 ([#2263](https://github.com/SemiAnalysisAI/InferenceX/pull/2263)), to save GPU cluster time for higher-priority real-world agentic-coding benchmarks and new frontier models. Archived configs lived in `configs/deprecated/` until it was deleted (#3464); see git history. The GLM-5.1 B200 TileRT point added later in [#2533](https://github.com/SemiAnalysisAI/InferenceX/pull/2533) remains active. |
 | Single-turn 1k8k | 1024 / 8192 | **Deprecated for all models** since 2026-03-27 ([#911](https://github.com/SemiAnalysisAI/InferenceX/pull/911)), to save GPU cluster time for higher-priority real-world agentic-coding benchmarks and new frontier models. Configs were removed, not archived. |
 
 ## AgentX Guidelines
@@ -176,7 +176,7 @@ Other offloading tiers, including NVMe KV cache offloading, are outside the init
 ## Notes
 
 - The `Prefix` column is the canonical `model-prefix` used in `configs/*-master.yaml` and by `generate_sweep_configs.py --model-prefix`.
-- "Retired" means the model no longer has any active scenario. Retired models' configs (except `llama70b`) are archived under [`configs/deprecated/`](configs/deprecated/).
+- "Retired" means the model no longer has any active scenario. Retired models' configs are deleted from the master configs; the former `configs/deprecated/` archive was removed in #3464, so past settings live in git history and `perf-changelog.yaml`.
 - Deprecating a precision (e.g. Qwen3.5 bf16) or one arm of an A/B pair (e.g. non-MTP) narrows a model's recipe coverage without retiring the model. The model stays listed as active as long as one scenario still runs.
 - `dsr1` began as the DeepSeek-V3 workflow templates in the initial repo import and was switched to DeepSeek-R1 benchmarking on 2025-08-13 (renamed `dsv3` → `dsr1` on 2025-08-20).
 - Adding a model? Follow [Add a model + hardware recipe](docs/configuration-procedures.md#add-a-model--hardware-recipe) and add a row here (and in [`MODELS_zh.md`](MODELS_zh.md)) in the same PR.
diff --git a/MODELS_zh.md b/MODELS_zh.md
index e2794deb5b..14143f9d5b 100644
--- a/MODELS_zh.md
+++ b/MODELS_zh.md
@@ -12,7 +12,7 @@ InferenceX-e2e 运行在数量固定且有限的 GPU 资源池上，并由一支
 
 **2026 年 8 月 3 日（星期一）**为下方第一张表中场景与精度的最后运行日，此后即告弃用。独立 A/B 基线的下线另见该表下方说明。
 
-**已于 2026 年 8 月 4 日部分执行**（[#2493](https://github.com/SemiAnalysisAI/InferenceX/pull/2493)）：第一张表中的场景与精度下线已完成。此次执行从启用的 master 配置中移除 54 个配置键并归档至 [`configs/deprecated/`](configs/deprecated/)，其基准测试脚本亦移入同级 `deprecated/` 目录。第二张表中的投机解码 A/B 下线**尚未执行**，详见该表下方说明。
+**已于 2026 年 8 月 4 日部分执行**（[#2493](https://github.com/SemiAnalysisAI/InferenceX/pull/2493)）：第一张表中的场景与精度下线已完成。此次执行从启用的 master 配置中移除 54 个配置键并归档至 `configs/deprecated/`，其基准测试脚本亦移入同级 `deprecated/` 目录。第二张表中的投机解码 A/B 下线**尚未执行**，详见该表下方说明。
 
 场景与精度下线：
 
@@ -42,7 +42,7 @@ InferenceX-e2e 运行在数量固定且有限的 GPU 资源池上，并由一支
 
 **2026 年 8 月 6 日（星期四）**为 **Kimi-K2.5/2.6/2.7-Code**（`kimik2.5`）**单轮 8k1k** 场景的最后运行日，此后该场景对这些模型弃用。原因：Kimi-K3 已于 2026 年 7 月 27 日发布，GPU 集群时间将转向更新的前沿模型。叠加上文的智能体编码弃用，`kimik2.5` 将不再有任何启用场景。该模型将于 **2026 年 8 月 6 日后完全退役**。
 
-**已于 2026-08-07 执行**（[#2527](https://github.com/SemiAnalysisAI/InferenceX/pull/2527)）：从启用的主配置中移除 17 个 `kimik2.5` 配置项，归档至 [`configs/deprecated/`](configs/deprecated/)，现分别合并至 `configs/deprecated/nvidia-master.yaml`（10 个）与 `configs/deprecated/amd-master.yaml`（7 个）。对应的 12 个基准测试脚本移入同级 `deprecated/` 目录。此后 `kimik2.5` 在所有主配置中**均无启用配置**，正式完全退役。同一 PR 还归档了 `kimik2.5-int4-h100-vllm`。#2493 将其脚本移入 `benchmarks/single_node/agentic/deprecated/` 时，该智能体编码配置项被遗留在 `nvidia-master.yaml` 中，现已与同类项一并归入 `configs/deprecated/nvidia-master.yaml`。SPEED-Bench 接受长度脚本 `benchmarks/single_node/speedbench/kimik2.5_fp4_b300_vllm.sh` 予以保留。Speedbench 由 `speedbench-al.yml` 驱动，不经过主配置，与 #2493 处理 MiniMax-M3 的方式一致。
+**已于 2026-08-07 执行**（[#2527](https://github.com/SemiAnalysisAI/InferenceX/pull/2527)）：从启用的主配置中移除 17 个 `kimik2.5` 配置项，归档至 `configs/deprecated/`，现分别合并至 `configs/deprecated/nvidia-master.yaml`（10 个）与 `configs/deprecated/amd-master.yaml`（7 个）。对应的 12 个基准测试脚本移入同级 `deprecated/` 目录。此后 `kimik2.5` 在所有主配置中**均无启用配置**，正式完全退役。同一 PR 还归档了 `kimik2.5-int4-h100-vllm`。#2493 将其脚本移入 `benchmarks/single_node/agentic/deprecated/` 时，该智能体编码配置项被遗留在 `nvidia-master.yaml` 中，现已与同类项一并归入 `configs/deprecated/nvidia-master.yaml`。SPEED-Bench 接受长度脚本 `benchmarks/single_node/speedbench/kimik2.5_fp4_b300_vllm.sh` 予以保留。Speedbench 由 `speedbench-al.yml` 驱动，不经过主配置，与 #2493 处理 MiniMax-M3 的方式一致。
 
 ### 2026 年 9 月 8 日（星期二）
 
@@ -54,9 +54,9 @@ InferenceX-e2e 运行在数量固定且有限的 GPU 资源池上，并由一支
 
 原因：`dsv4` 是本仓库中单轮场景占用最大的模型。当前有 45 个启用的配置项使用 8k1k 场景（`configs/nvidia-master.yaml` 32 个，`configs/amd-master.yaml` 13 个），覆盖 H200、B200、B300、GB200、GB300、MI300X、MI325X 与 MI355X，涉及 vLLM、SGLang、TensorRT-LLM、ATOM、Dynamo 与 llm-d，在每一轮完整 sweep 中占比可观。AgentX 轨迹回放才是 AI 实验室与 ML 社区真正关注的场景，而 DeepSeek-V4-Pro 的 19 个智能体编码配置项正是 `dsv4` 中支撑已发布北极星（North Star）帕累托前沿的部分。下线固定序列长度分支可为 AgentX 以及 Qwen3.8-Flash-Next 等新前沿模型腾出集群机时，同时不减少该模型对外发布的内容。对于仍列有该场景的其他模型，单轮 8k1k 保持启用。
 
-**已于 2026-09-09 执行**（[#2921](https://github.com/SemiAnalysisAI/InferenceX/pull/2921)）：46 个 `dsv4` 8k1k 配置项已从启用的主配置中移除并归档至 [`configs/deprecated/`](configs/deprecated/)，现合并至 `configs/deprecated/nvidia-master.yaml`（33 个）与 `configs/deprecated/amd-master.yaml`（13 个）；对应的 28 个基准测试脚本移入同级 `deprecated/` 目录（`benchmarks/single_node/fixed_seq_len/` 下 25 个，`benchmarks/multi_node/` 下 3 个），与 [#2493](https://github.com/SemiAnalysisAI/InferenceX/pull/2493) 和 [#2527](https://github.com/SemiAnalysisAI/InferenceX/pull/2527) 的做法一致。数量为 46 而非上文所述的 45，是因为 `dsv4-fp4-b200-dynamo-sglang` 在本公告发布后才合入。19 个智能体编码配置项未做改动：`dsv4` 以智能体编码为唯一场景继续运行与发布。`dsv4` 的 SPEED-Bench 接受长度脚本予以保留。Speedbench 由 `speedbench-al.yml` 驱动，不经过主配置。已归档多节点配置项所引用的 srt-slurm 与 llm-d 配方 YAML 作为惰性参考数据原地保留，与 #2493 和 #2527 的处理一致。
+**已于 2026-09-09 执行**（[#2921](https://github.com/SemiAnalysisAI/InferenceX/pull/2921)）：46 个 `dsv4` 8k1k 配置项已从启用的主配置中移除并归档至 `configs/deprecated/`，现合并至 `configs/deprecated/nvidia-master.yaml`（33 个）与 `configs/deprecated/amd-master.yaml`（13 个）；对应的 28 个基准测试脚本移入同级 `deprecated/` 目录（`benchmarks/single_node/fixed_seq_len/` 下 25 个，`benchmarks/multi_node/` 下 3 个），与 [#2493](https://github.com/SemiAnalysisAI/InferenceX/pull/2493) 和 [#2527](https://github.com/SemiAnalysisAI/InferenceX/pull/2527) 的做法一致。数量为 46 而非上文所述的 45，是因为 `dsv4-fp4-b200-dynamo-sglang` 在本公告发布后才合入。19 个智能体编码配置项未做改动：`dsv4` 以智能体编码为唯一场景继续运行与发布。`dsv4` 的 SPEED-Bench 接受长度脚本予以保留。Speedbench 由 `speedbench-al.yml` 驱动，不经过主配置。已归档多节点配置项所引用的 srt-slurm 与 llm-d 配方 YAML 作为惰性参考数据原地保留，与 #2493 和 #2527 的处理一致。
 
-**弃用状态一致性核查（2026-09-21）：** 启用的主配置及基准测试脚本位置与上述已执行的退役事项和下方支持矩阵一致。GLM-5.1 B200 TileRT 仍是文档明确保留的例外，不受此前 GLM-5/5.1 和 1k1k 退役范围限制。有条件的 A/B 基线退役仍待执行；对 Pareto 前沿有贡献的非投机解码配置继续受支持。进一步的路由核查还移除了启动器和运行时设置中遗留的退役模型分支及 GLM-5 专用环境覆盖，并修正了仍推荐退役配置的工作流和智能体指南。SPEED-Bench 采集器、历史结果读取逻辑及明确保留的配方 YAML 继续保留。弃用配置现统一归档至 [`configs/deprecated/amd-master.yaml`](configs/deprecated/amd-master.yaml) 和 [`configs/deprecated/nvidia-master.yaml`](configs/deprecated/nvidia-master.yaml)。
+**弃用状态一致性核查（2026-09-21）：** 启用的主配置及基准测试脚本位置与上述已执行的退役事项和下方支持矩阵一致。GLM-5.1 B200 TileRT 仍是文档明确保留的例外，不受此前 GLM-5/5.1 和 1k1k 退役范围限制。有条件的 A/B 基线退役仍待执行；对 Pareto 前沿有贡献的非投机解码配置继续受支持。进一步的路由核查还移除了启动器和运行时设置中遗留的退役模型分支及 GLM-5 专用环境覆盖，并修正了仍推荐退役配置的工作流和智能体指南。SPEED-Bench 采集器、历史结果读取逻辑及明确保留的配方 YAML 继续保留。弃用配置曾统一归档至 `configs/deprecated/amd-master.yaml` 和 `configs/deprecated/nvidia-master.yaml`。
 
 **单节点切换为仅使用 SRT（2026-09-22）：** 活跃的单节点定长配方现统一使用 SRT-Slurm。两个仅支持 Docker 的 Qwen3.5 RTX PRO 6000 FP4 配置（启用和关闭 MTP）已退役，原始设置保留在 `configs/deprecated/nvidia-master.yaml`，脚本保留在 `benchmarks/single_node/fixed_seq_len/deprecated/`。已移除不再使用的 `rtx6000pro-lat` runner 映射、启动器及运行时设置。Qwen3.5 在其他受支持的 Slurm 池上继续启用；AgentX 和多节点覆盖保持不变。
 
@@ -66,7 +66,7 @@ InferenceX-e2e 运行在数量固定且有限的 GPU 资源池上，并由一支
 |---|---|---|
 | 智能体编码（agentic coding） | 长上下文、多轮真实流量的轨迹回放，含子智能体（sub agents） | 启用。此场景采用基于轨迹回放的智能体编码基准测试（见 [`benchmarks/single_node/srt-slurm-recipes/`](benchmarks/single_node/srt-slurm-recipes/) 下的 srt-slurm 配方与共享客户端 [`benchmarks/srt_agentic.sh`](benchmarks/srt_agentic.sh)）。今后新模型预计将仅以智能体编码场景接入。可开启或关闭投机解码以获得最优帕累托点；不要求独立的非投机解码 A/B 基线（见[弃用公告](#弃用公告)）。 |
 | 单轮 8k1k | 8192 / 1024 | 启用。当前主要的固定序列长度（fixed-seq-len）场景。 |
-| 单轮 1k1k | 1024 / 1024 | 自 2026-07-17 起弃用（[#2263](https://github.com/SemiAnalysisAI/InferenceX/pull/2263)），以便将 GPU 集群时间留给优先级更高的真实场景智能体编码基准测试与新的前沿模型。归档配置位于 [`configs/deprecated/`](configs/deprecated/)。后续由 [#2533](https://github.com/SemiAnalysisAI/InferenceX/pull/2533) 加入的 GLM-5.1 B200 TileRT 测试点仍启用。 |
+| 单轮 1k1k | 1024 / 1024 | 自 2026-07-17 起弃用（[#2263](https://github.com/SemiAnalysisAI/InferenceX/pull/2263)），以便将 GPU 集群时间留给优先级更高的真实场景智能体编码基准测试与新的前沿模型。归档配置曾位于 `configs/deprecated/`，该目录已在 #3464 中删除，请查阅 Git 历史。后续由 [#2533](https://github.com/SemiAnalysisAI/InferenceX/pull/2533) 加入的 GLM-5.1 B200 TileRT 测试点仍启用。 |
 | 单轮 1k8k | 1024 / 8192 | **对所有模型均已弃用**，自 2026-03-27 起（[#911](https://github.com/SemiAnalysisAI/InferenceX/pull/911)），以便将 GPU 集群时间留给优先级更高的真实场景智能体编码基准测试与新的前沿模型。相关配置已删除，未归档。 |
 
 ## AgentX 指南
@@ -176,7 +176,7 @@ InferenceX 支持 SGLang 和 vLLM 双方的维护者，并响应 AI 实验室和
 ## 说明
 
 - 「前缀」列为 `configs/*-master.yaml` 中的规范 `model-prefix`，同时用于 `generate_sweep_configs.py --model-prefix`。
-- 「退役」指该模型已无任何启用场景。退役模型的配置（`llama70b` 除外）归档于 [`configs/deprecated/`](configs/deprecated/)。
+- 「退役」指该模型已无任何启用场景。退役模型的配置直接从主配置中删除；原 `configs/deprecated/` 归档目录已在 #3464 中删除，历史设置保留在 Git 历史与 `perf-changelog.yaml` 中。
 - 弃用某一精度（如 Qwen3.5 bf16）或 A/B 对照中的某一分支（如非 MTP），只是收窄该模型的配方覆盖范围，并不等于模型退役；只要仍有一个场景在运行，该模型即继续列为启用状态。
 - `dsr1` 最初以 DeepSeek-V3 workflow 模板的形式随仓库首次导入，2025-08-13 切换为 DeepSeek-R1 基准测试（2025-08-20 将 `dsv3` 重命名为 `dsr1`）。
 - 新增模型时，请按[添加模型 + 硬件配方](docs/configuration-procedures_zh.md#添加模型--硬件配方)流程操作，并在同一 PR 中同时更新本文件与 [`MODELS.md`](MODELS.md) 的表格。
diff --git a/benchmarks/multi_node/amd_utils/deprecated/models_vllm.yaml b/benchmarks/multi_node/amd_utils/deprecated/models_vllm.yaml
index 4ee11e7e77..f9a54d918a 100644
--- a/benchmarks/multi_node/amd_utils/deprecated/models_vllm.yaml
+++ b/benchmarks/multi_node/amd_utils/deprecated/models_vllm.yaml
@@ -35,8 +35,8 @@ MiniMax-M3-MXFP8:
   # varying the prefill:decode worker ratio rather than TP).
   # --block-size 128 is mandatory (MSA sparse/index cache); text-only benchmark
   # so --language-model-only frees the vision encoder. gfx950 uses FP8 KV cache.
-  # Serve flags + env kept in sync with the single-node recipe
-  # benchmarks/single_node/fixed_seq_len/deprecated/minimaxm3_fp8_mi355x.sh:
+  # Serve flags + env were kept in sync with the former single-node
+  # minimaxm3_fp8_mi355x.sh recipe (since deleted):
   #   --moe-backend aiter (AITER fused MoE), --linear-backend emulation, and the
   #   larger --max-num-batched-tokens 32768 prefill budget; plus the AITER
   #   router-append shared-experts fusion (self-disables under EP; the sweep is
diff --git a/configs/deprecated/amd-master.yaml b/configs/deprecated/amd-master.yaml
deleted file mode 100644
index 0bf2a275e1..0000000000
--- a/configs/deprecated/amd-master.yaml
+++ /dev/null
@@ -1,3464 +0,0 @@
-# Deprecated AMD configurations; historical reference only.
-# Active configs live in configs/amd-master.yaml.
-# Colliding historical 1k1k keys have a -deprecated-1k1k suffix;
-# their original keys are recorded in comments. Preserve each archived version.
-
-# Archive source: amd-1k1k-master.yaml
-# Deprecated 1k1k fixed-sequence entries archived from amd-master.yaml.
-# Removed from the active master config so sweep generation no longer selects them.
-
-dsr1-fp4-mi355x-sglang:
-  image: lmsysorg/sglang:v0.5.12-rocm700-mi35x
-  model: amd/DeepSeek-R1-0528-MXFP4-Preview
-  model-prefix: dsr1
-  runner: mi355x
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 4, conc-end: 64 }
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-dsr1-fp4-mi355x-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.12-rocm700-mi35x
-  model: amd/DeepSeek-R1-0528-MXFP4
-  model-prefix: dsr1
-  runner: mi355x
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-
-dsr1-fp4-mi355x-atom:
-  image: rocm/atom:rocm7.2.3_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom20260511
-  model: amd/DeepSeek-R1-0528-MXFP4-Preview
-  model-prefix: dsr1
-  runner: mi355x
-  precision: fp4
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, ep: 1, conc-start: 32, conc-end: 256 }
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 32 }
-
-dsr1-fp4-mi355x-atom-mtp:
-  image: rocm/atom:rocm7.2.3_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom20260511
-  model: amd/DeepSeek-R1-0528-MXFP4
-  model-prefix: dsr1
-  runner: mi355x
-  precision: fp4
-  # WIP framework (no customers yet)
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 4, conc-end: 256, spec-decoding: mtp }
-      - { tp: 8, conc-start: 4, conc-end: 256, spec-decoding: mtp }
-
-dsr1-fp8-mi300x-sglang:
-  image: lmsysorg/sglang:v0.5.12-rocm700-mi30x
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: mi300x
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-dsr1-fp8-mi325x-sglang:
-  image: lmsysorg/sglang:v0.5.12-rocm700-mi30x
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: mi325x
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-# Original config key: dsv4-fp8-mi300x-vllm
-dsv4-fp8-mi300x-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:nightly-09663abde0f50944a8d5ea30120666024b503faa
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: mi300x
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 512 }
-
-# Original config key: dsv4-fp8-mi300x-vllm-mtp
-dsv4-fp8-mi300x-vllm-mtp-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:nightly-09663abde0f50944a8d5ea30120666024b503faa
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: mi300x
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 512, spec-decoding: mtp }
-
-dsr1-fp8-mi355x-sglang:
-  image: lmsysorg/sglang:v0.5.12-rocm700-mi35x
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: mi355x
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-dsr1-fp8-mi355x-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.12-rocm700-mi35x
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: mi355x
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-
-# Original config key: qwen3.5-bf16-mi355x-sglang
-qwen3.5-bf16-mi355x-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang-rocm:v0.5.12-rocm720-mi35x-20260517
-  model: Qwen/Qwen3.5-397B-A17B
-  model-prefix: qwen3.5
-  runner: mi355x
-  precision: bf16
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 256 }
-
-# Original config key: qwen3.5-bf16-mi355x-sglang-mtp
-qwen3.5-bf16-mi355x-sglang-mtp-deprecated-1k1k:
-  image: lmsysorg/sglang-rocm:v0.5.12-rocm720-mi35x-20260517
-  model: Qwen/Qwen3.5-397B-A17B
-  model-prefix: qwen3.5
-  runner: mi355x
-  precision: bf16
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 256, spec-decoding: mtp }
-
-# Original config key: qwen3.5-bf16-mi300x-sglang
-qwen3.5-bf16-mi300x-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.12-rocm720-mi30x
-  model: Qwen/Qwen3.5-397B-A17B
-  model-prefix: qwen3.5
-  runner: mi300x
-  precision: bf16
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-# Original config key: qwen3.5-bf16-mi325x-sglang
-qwen3.5-bf16-mi325x-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.12-rocm720-mi30x
-  model: Qwen/Qwen3.5-397B-A17B
-  model-prefix: qwen3.5
-  runner: mi325x
-  precision: bf16
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-qwen3.5-fp8-mi325x-sglang:
-  image: lmsysorg/sglang:v0.5.12-rocm720-mi30x
-  model: Qwen/Qwen3.5-397B-A17B-FP8
-  model-prefix: qwen3.5
-  runner: mi325x
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-qwen3.5-fp8-mi355x-sglang:
-  image: lmsysorg/sglang:v0.5.14-rocm720-mi35x
-  model: Qwen/Qwen3.5-397B-A17B-FP8
-  model-prefix: qwen3.5
-  runner: mi355x
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 }
-
-qwen3.5-fp8-mi355x-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.14-rocm720-mi35x
-  model: Qwen/Qwen3.5-397B-A17B-FP8
-  model-prefix: qwen3.5
-  runner: mi355x
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 256, spec-decoding: mtp }
-
-qwen3.5-fp8-mi355x-atom:
-  image: rocm/atom:rocm7.2.3_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom20260511
-  model: Qwen/Qwen3.5-397B-A17B-FP8
-  model-prefix: qwen3.5
-  runner: mi355x
-  precision: fp8
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 2, ep: 1, conc-start: 4, conc-end: 256 }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 }
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 256 }
-
-qwen3.5-fp8-mi355x-atom-mtp:
-  image: rocm/atom:rocm7.2.2_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom0.1.2.post
-  model: Qwen/Qwen3.5-397B-A17B-FP8
-  model-prefix: qwen3.5
-  runner: mi355x
-  precision: fp8
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 256, spec-decoding: mtp }
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 256, spec-decoding: mtp }
-
-qwen3.5-fp8-mi355x-sglang-disagg:
-  image: lmsysorg/sglang:v0.5.14-rocm720-mi35x
-  model: Qwen/Qwen3.5-397B-A17B-FP8
-  model-prefix: qwen3.5
-  runner: cluster:mi355x-amds
-  precision: fp8
-  framework: sglang-disagg
-  router: { name: sglang-router, version: "0.3.2" }
-  kv-p2p-transfer: mori
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # Matches qwen3.5-fp8-mi355x-sglang TP8/EP1 low-concurrency sweep
-      - spec-decoding: "none"
-        conc-list: [ 8, 16, 32, 64, 128 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=0"
-
-qwen3.5-fp4-mi355x-sglang:
-  image: lmsysorg/sglang-rocm:v0.5.15-rocm720-mi35x-20260713
-  model: amd/Qwen3.5-397B-A17B-MXFP4
-  model-prefix: qwen3.5
-  runner: mi355x
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 2, conc-start: 4, conc-end: 256 }
-      - { tp: 4, conc-start: 4, conc-end: 16 }
-
-qwen3.5-fp4-mi355x-atom:
-  image: rocm/atom:rocm7.2.2_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom0.1.2.post
-  model: amd/Qwen3.5-397B-A17B-MXFP4
-  model-prefix: qwen3.5
-  runner: mi355x
-  precision: fp4
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 2, conc-start: 4, conc-end: 256 }
-      - { tp: 4, conc-start: 4, conc-end: 16 }
-
-qwen3.5-fp4-mi355x-sglang-mtp:
-  image: lmsysorg/sglang-rocm:v0.5.15-rocm720-mi35x-20260713
-  model: amd/Qwen3.5-397B-A17B-MXFP4
-  model-prefix: qwen3.5
-  runner: mi355x
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 2, conc-start: 4, conc-end: 128, spec-decoding: mtp }
-      - { tp: 4, conc-start: 4, conc-end: 16, spec-decoding: mtp }
-
-qwen3.5-fp4-mi355x-sglang-disagg:
-  image: lmsysorg/sglang-rocm:v0.5.12.post1-rocm720-mi35x-20260523
-  model: amd/Qwen3.5-397B-A17B-MXFP4
-  model-prefix: qwen3.5
-  runner: cluster:mi355x-amds
-  precision: fp4
-  framework: sglang-disagg
-  router: { name: sglang-router, version: "0.3.2" }
-  kv-p2p-transfer: mori
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 1P1D TP8/EP1, dp-attn false; MoRI conn.py overlay via job.slurm.
-      - spec-decoding: "none"
-        conc-list: [ 8, 16, 32, 64, 128, 256, 512 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=0"
-
-qwen3.5-fp8-mi300x-sglang:
-  image: lmsysorg/sglang:v0.5.12-rocm720-mi30x
-  model: Qwen/Qwen3.5-397B-A17B-FP8
-  model-prefix: qwen3.5
-  runner: mi300x
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-# Original config key: glm5-fp8-mi355x-sglang
-glm5-fp8-mi355x-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang-rocm:v0.5.12-rocm720-mi35x-20260517
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: mi355x
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 4, conc-end: 256 }
-
-# Original config key: glm5-fp8-mi355x-sglang-mtp
-glm5-fp8-mi355x-sglang-mtp-deprecated-1k1k:
-  image: lmsysorg/sglang-rocm:v0.5.12-rocm720-mi35x-20260517
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: mi355x
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 4, conc-end: 128, spec-decoding: mtp }
-      - { tp: 8, conc-start: 4, conc-end: 8, spec-decoding: mtp }
-
-# Original config key: glm5-fp8-mi355x-sglang-disagg
-glm5-fp8-mi355x-sglang-disagg-deprecated-1k1k:
-  image: lmsysorg/sglang-rocm:v0.5.12.post1-rocm720-mi35x-20260523
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: cluster:mi355x-amds
-  precision: fp8
-  framework: sglang-disagg
-  router: { name: sglang-router, version: "0.3.2" }
-  kv-p2p-transfer: mori
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 1P+1D TP8/EP1 CI smoke sweep (aligned with glm5-fp8-mi355x-sglang conc range)
-      - spec-decoding: "none"
-        conc-list: [ 8, 16, 32, 64, 128, 256, 512 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=0"
-
-# Original config key: glm5-fp8-mi355x-atom
-glm5-fp8-mi355x-atom-deprecated-1k1k:
-  image: rocm/atom:rocm7.2.2_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom0.1.2.post
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: mi355x
-  precision: fp8
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 4, conc-end: 256 }
-      - { tp: 8, conc-start: 4, conc-end: 256 }
-
-# Original config key: glm5.1-fp4-mi355x-sglang
-glm5.1-fp4-mi355x-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang-rocm:v0.5.12.post1-rocm720-mi35x-20260529
-  model: amd/GLM-5.1-MXFP4
-  model-prefix: glm5.1
-  runner: mi355x
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 2, conc-start: 4, conc-end: 256 }
-      - { tp: 4, conc-start: 4, conc-end: 16 }
-
-# Original config key: glm5.1-fp4-mi355x-atom
-glm5.1-fp4-mi355x-atom-deprecated-1k1k:
-  image: rocm/atom:rocm7.2.2_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom0.1.2.post
-  model: amd/GLM-5.1-MXFP4
-  model-prefix: glm5.1
-  runner: mi355x
-  precision: fp4
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 4, conc-end: 256 }
-
-# Original config key: kimik2.5-int4-mi355x-vllm
-kimik2.5-int4-mi355x-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:v0.24.0
-  model: moonshotai/Kimi-K2.5
-  model-prefix: kimik2.5
-  runner: mi355x
-  precision: int4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 128 }
-      - { tp: 4, conc-start: 4, conc-end: 128 }
-
-# Original config key: kimik2.5-int4-mi325x-vllm
-kimik2.5-int4-mi325x-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:v0.21.0
-  model: moonshotai/Kimi-K2.5
-  model-prefix: kimik2.5
-  runner: mi325x
-  precision: int4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-# Original config key: kimik2.5-int4-mi300x-vllm
-kimik2.5-int4-mi300x-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:v0.21.0
-  model: moonshotai/Kimi-K2.5
-  model-prefix: kimik2.5
-  runner: mi300x
-  precision: int4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-# Original config key: kimik2.5-fp4-mi355x-vllm
-kimik2.5-fp4-mi355x-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:v0.24.0
-  model: amd/Kimi-K2.5-MXFP4
-  model-prefix: kimik2.5
-  runner: mi355x
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-      - { tp: 4, conc-start: 4, conc-end: 64 }
-
-# Original config key: kimik2.5-fp4-mi355x-atom
-kimik2.5-fp4-mi355x-atom-deprecated-1k1k:
-  image: rocm/atom:rocm7.2.4_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom0.1.4_202607091539
-  model: amd/Kimi-K2.5-MXFP4
-  model-prefix: kimik2.5
-  runner: mi355x
-  precision: fp4
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 4, conc-end: 128 }
-
-dsr1-fp8-mi355x-atom:
-  image: rocm/atom:rocm7.2.3_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom20260511
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: mi355x
-  precision: fp8
-  # WIP framework (no customers yet)
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 128 }
-
-dsr1-fp8-mi355x-atom-mtp:
-  image: rocm/atom:rocm7.2.4_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom0.1.3
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: mi355x
-  precision: fp8
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 512, spec-decoding: mtp }
-
-dsr1-fp8-mi355x-sglang-disagg:
-  image: rocm/sgl-dev:sglang-0.5.9-rocm720-mi35x-mori-0227-2
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: cluster:mi355x-amds
-  precision: fp8
-  framework: sglang-disagg
-  router: { name: sglang-router, version: "0.3.2" }
-  kv-p2p-transfer: mori
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # non-MTP configurations
-      # "Top of curve" (1 prefill workers each at DEP8 and 1 decode workers at DEP16)
-      - spec-decoding: "none"
-        conc-list: [ 1024, 2048 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "DECODE_MTP_SIZE=0"
-
-      # "Middle of curve" (1 prefill workers each at TP8 and 2 decode workers at DEP8)
-      - spec-decoding: "none"
-        conc-list: [ 1536, 1024, 512 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "DECODE_MTP_SIZE=0"
-
-      # "Bottom of curve" (1 prefill worker at TEP8 and 2 decode workers at TEP8)
-      - spec-decoding: "none"
-        conc-list: [ 256, 128, 64, 32, 16, 8, 4 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "DECODE_MTP_SIZE=0"
-
-      - spec-decoding: "none"
-        conc-list: [ 64, 32, 16, 8, 4, 2, 1 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=0"
-
-dsr1-fp8-mi355x-sglang-disagg-mtp:
-  image: rocm/sgl-dev:sglang-0.5.9-rocm720-mi35x-mori-0227-2
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: cluster:mi355x-amds
-  precision: fp8
-  framework: sglang-disagg
-  router: { name: sglang-router, version: "0.3.2" }
-  kv-p2p-transfer: mori
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # MTP configurations
-      # "Top of curve" (1 prefill worker at DEP8 and 1 decode worker at DEP16)
-      - spec-decoding: "mtp"
-        conc-list: [ 1024, 2048 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "DECODE_MTP_SIZE=1"
-
-      # "Middle of curve" (1 prefill worker at TP8 and 2 decode workers each at DEP8)
-      - spec-decoding: "mtp"
-        conc-list: [ 1536, 1024, 512, 256 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "DECODE_MTP_SIZE=1"
-
-      # "Bottom of curve" (1 prefill worker at TEP8 and 2 decode workers at TEP8)
-      - spec-decoding: "mtp"
-        conc-list: [ 256, 128, 64, 32, 16, 8, 4 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "DECODE_MTP_SIZE=2"
-
-      - spec-decoding: "mtp"
-        conc-list: [ 64, 32, 16, 8, 4, 2, 1 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=2"
-
-# Original config key: kimik2.5-fp4-mi355x-vllm-disagg
-kimik2.5-fp4-mi355x-vllm-disagg-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:v0.24.0
-  model: amd/Kimi-K2.5-MXFP4
-  model-prefix: kimik2.5
-  runner: cluster:mi355x-amds
-  precision: fp4
-  framework: vllm-disagg
-  router: { name: vllm-router, version: "0.1.14" }
-  kv-p2p-transfer: moriio
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 1P2D: 1 prefill node (co-located with proxy) + 2 decode nodes = 3 nodes total
-      - spec-decoding: "none"
-        conc-list: [ 8, 16, 32, 64, 128, 256, 512 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-
-dsr1-fp4-mi355x-sglang-disagg:
-  image: lmsysorg/sglang-rocm:v0.5.12-rocm720-mi35x-20260519
-  model: amd/DeepSeek-R1-0528-MXFP4-v2
-  model-prefix: dsr1
-  runner: cluster:mi355x-amds
-  precision: fp4
-  framework: sglang-disagg
-  router: { name: sglang-router, version: "0.3.2" }
-  kv-p2p-transfer: mori
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # non-MTP configurations
-      # 1P1D TP8
-      - spec-decoding: "none"
-        conc-list: [ 1, 2, 4, 8 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=0"
-
-      # 1P2D TP8
-      - spec-decoding: "none"
-        conc-list: [ 2, 4, 8, 16, 32 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "DECODE_MTP_SIZE=0"
-
-      # 1P2D TP8
-      - spec-decoding: "none"
-        conc-list: [ 64, 128, 256 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "DECODE_MTP_SIZE=0"
-
-      # 1P2D TP4
-      - spec-decoding: "none"
-        conc-list: [ 64, 128, 256 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "DECODE_MTP_SIZE=0"
-
-      # 1*DEP4+ 1*DEP8
-      - spec-decoding: "none"
-        conc-list: [ 1024, 2048, 4096 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=0"
-
-dsr1-fp4-mi355x-sglang-disagg-1k1k-mtp:
-  image: lmsysorg/sglang-rocm:v0.5.12.post1-rocm720-mi35x-20260529
-  model: amd/DeepSeek-R1-0528-MXFP4-v2
-  model-prefix: dsr1
-  runner: cluster:mi355x-amds
-  precision: fp4
-  framework: sglang-disagg
-  router: { name: sglang-router, version: "0.3.2" }
-  kv-p2p-transfer: mori
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # MTP configurations
-      # 1P1D TP8
-      - spec-decoding: "mtp"
-        conc-list: [ 1, 2, 4, 8 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=3"
-
-      # 1P2D TP8
-      - spec-decoding: "mtp"
-        conc-list: [ 2, 4, 8, 16, 32 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "DECODE_MTP_SIZE=3"
-
-      # 1P2D TP8
-      - spec-decoding: "mtp"
-        conc-list: [ 64, 128, 256 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "DECODE_MTP_SIZE=2"
-
-      # 1P2D TP4
-      - spec-decoding: "mtp"
-        conc-list: [ 64, 128, 256 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "DECODE_MTP_SIZE=2"
-
-      # 1*DEP4+ 1*DEP8
-      - spec-decoding: "mtp"
-        conc-list: [ 1024, 2048, 4096 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=1"
-
-# Original config key: dsv4-fp4-mi355x-sglang
-dsv4-fp4-mi355x-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang-rocm:v0.5.14-rocm720-mi35x-20260706
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: mi355x
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, dp-attn: true, conc-start: 64, conc-end: 2048 }
-      - { tp: 4, dp-attn: true, conc-start: 16, conc-end: 128 }
-      - { tp: 4, dp-attn: false, conc-start: 1 , conc-end: 32 }
-
-# Original config key: dsv4-fp4-mi355x-sglang-mtp
-dsv4-fp4-mi355x-sglang-mtp-deprecated-1k1k:
-  image: lmsysorg/sglang-rocm:v0.5.14-rocm720-mi35x-20260708
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: mi355x
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, dp-attn: true, conc-start: 64, conc-end: 2048, spec-decoding: mtp }
-      - { tp: 8, dp-attn: false, conc-start: 1, conc-end: 32, spec-decoding: mtp }
-
-# Original config key: dsv4-fp4-mi355x-vllm
-dsv4-fp4-mi355x-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:nightly-09663abde0f50944a8d5ea30120666024b503faa
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: mi355x
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 512 }
-
-# Original config key: dsv4-fp4-mi355x-vllm-mtp
-dsv4-fp4-mi355x-vllm-mtp-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:nightly-09663abde0f50944a8d5ea30120666024b503faa
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: mi355x
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 512, spec-decoding: mtp }
-
-# Original config key: dsv4-fp4-mi355x-atom
-dsv4-fp4-mi355x-atom-deprecated-1k1k:
-  image: rocm/atom-dev:nightly_202606161823
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: mi355x
-  precision: fp4
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-        # conc4-64, TP8
-        # conc128-512, DPA
-        # conc1024-2048, DPA TBO
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 64 }
-      - { tp: 8, ep: 1, dp-attn: true, conc-start: 64, conc-end: 2048 }
-
-# Original config key: dsv4-fp4-mi355x-atom-mtp
-dsv4-fp4-mi355x-atom-mtp-deprecated-1k1k:
-  image: rocm/atom:rocm7.2.4_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom0.1.3
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: mi355x
-  precision: fp4
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 1024, spec-decoding: mtp }
-
-# Original config key: qwen3.5-bf16-mi325x-sglang-mtp
-qwen3.5-bf16-mi325x-sglang-mtp-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.12-rocm720-mi30x
-  model: Qwen/Qwen3.5-397B-A17B
-  model-prefix: qwen3.5
-  runner: mi325x
-  precision: bf16
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-
-dsr1-fp8-mi325x-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.12-rocm700-mi30x
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: mi325x
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-
-qwen3.5-fp8-mi325x-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.12-rocm720-mi30x
-  model: Qwen/Qwen3.5-397B-A17B-FP8
-  model-prefix: qwen3.5
-  runner: mi325x
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-
-# Original config key: glm5-fp8-mi325x-sglang
-glm5-fp8-mi325x-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.12-rocm720-mi30x
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: mi325x
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-# Original config key: glm5-fp8-mi325x-sglang-mtp
-glm5-fp8-mi325x-sglang-mtp-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.12-rocm720-mi30x
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: mi325x
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-
-dsr1-fp4-mi355x-sglang-disagg-mtp:
-  image: lmsysorg/sglang-rocm:v0.5.12-rocm720-mi35x-20260519
-  model: amd/DeepSeek-R1-0528-MXFP4-v2
-  model-prefix: dsr1
-  runner: cluster:mi355x-amds
-  precision: fp4
-  framework: sglang-disagg
-  router: { name: sglang-router, version: "0.3.2" }
-  kv-p2p-transfer: mori
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # MTP configurations
-      # 1P1D TP8
-      - spec-decoding: "mtp"
-        conc-list: [ 1, 2, 4, 8 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=3"
-
-      # 1P2D TP8
-      - spec-decoding: "mtp"
-        conc-list: [ 2, 4, 8, 16, 32 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "DECODE_MTP_SIZE=3"
-
-      # 1P2D TP8
-      - spec-decoding: "mtp"
-        conc-list: [ 64, 128, 256 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "DECODE_MTP_SIZE=2"
-
-      # 1P2D TP4
-      - spec-decoding: "mtp"
-        conc-list: [ 64, 128, 256 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "DECODE_MTP_SIZE=2"
-
-      # 1*DEP4+ 1*DEP8
-      - spec-decoding: "mtp"
-        conc-list: [ 1024, 2048, 4096 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=1"
-
-# Original config key: dsv4-fp4-mi355x-atom-disagg
-dsv4-fp4-mi355x-atom-disagg-deprecated-1k1k:
-  image: rocm/atom-dev:nightly_202606101403
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: mi355x
-  precision: fp4
-  framework: atom-disagg
-  router: { name: atomesh, version: "087b82d9c1f630e79149ba37e6213257ec9a76f8" }
-  kv-p2p-transfer: mooncake
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - conc-list: [ 4, 8, 16, 32, 64, 128, 256, 512, 1024 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-
-# MiniMax-M3 MXFP8 MI355X recipe:
-# https://github.com/vllm-project/recipes/commit/2a3728ed9892debfd767a72a58ebc90b33f186e5
-# MXFP8 runs from TP=4 on gfx950; block size 128 is mandatory for MSA.
-# Original config key: minimaxm3-fp8-mi355x-vllm
-minimaxm3-fp8-mi355x-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:nightly-09663abde0f50944a8d5ea30120666024b503faa
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: mi355x
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 512 }
-
-# Original config key: minimaxm3-fp8-mi355x-vllm-mtp
-minimaxm3-fp8-mi355x-vllm-mtp-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:nightly-09663abde0f50944a8d5ea30120666024b503faa
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: mi355x
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 512, spec-decoding: mtp }
-
-# Original config key: minimaxm3-fp4-mi355x-vllm
-minimaxm3-fp4-mi355x-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:nightly-4559c43a9526597c00cbcc4f59979496500268d1
-  model: amd/MiniMax-M3-MXFP4
-  model-prefix: minimaxm3
-  runner: mi355x
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 64 }
-      - { tp: 8, ep: 8, conc-start: 1, conc-end: 512 }
-      - { tp: 4, conc-start: 1, conc-end: 64 }
-      - { tp: 4, ep: 4, conc-start: 64, conc-end: 512 }
-      - { tp: 2, ep: 2, conc-start: 16, conc-end: 128 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 1024 }
-
-# Original config key: minimaxm3-fp4-mi355x-vllm-mtp
-minimaxm3-fp4-mi355x-vllm-mtp-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:nightly-3f5a1e1733200760169ff31ebe60a271072b199e
-  model: amd/MiniMax-M3-MXFP4
-  model-prefix: minimaxm3
-  runner: mi355x
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 8, ep: 8, conc-start: 1, conc-end: 256, spec-decoding: mtp }
-      - { tp: 4, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 4, ep: 4, conc-start: 64, conc-end: 256, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 512, spec-decoding: mtp }
-
-# Original config key: minimaxm3-fp4-mi355x-atom
-minimaxm3-fp4-mi355x-atom-deprecated-1k1k:
-  image: rocm/atom-dev:nightly_202607011530
-  model: amd/MiniMax-M3-MXFP4
-  model-prefix: minimaxm3
-  runner: mi355x
-  precision: fp4
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 256 }
-
-# Original config key: minimaxm3-fp4-mi355x-atom-mtp
-minimaxm3-fp4-mi355x-atom-mtp-deprecated-1k1k:
-  image: rocm/atom-dev:nightly_202607011530
-  model: amd/MiniMax-M3-MXFP4
-  model-prefix: minimaxm3
-  runner: mi355x
-  precision: fp4
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 256, spec-decoding: mtp  }
-
-# Original config key: minimaxm3-fp8-mi355x-atom
-minimaxm3-fp8-mi355x-atom-deprecated-1k1k:
-  image: rocm/atom-dev:nightly_202607011530
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: mi355x
-  precision: fp8
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 256 }
-
-# Original config key: minimaxm3-fp8-mi355x-atom-mtp
-minimaxm3-fp8-mi355x-atom-mtp-deprecated-1k1k:
-  image: rocm/atom-dev:nightly_202607011530
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: mi355x
-  precision: fp8
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 256, spec-decoding: mtp }
-
-# Original config key: minimaxm3-fp8-mi355x-atom-disagg
-minimaxm3-fp8-mi355x-atom-disagg-deprecated-1k1k:
-  image: rocm/atom-dev:nightly_202607011530
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: cluster:mi355x-amds
-  precision: fp8
-  framework: atom-disagg
-  router: { name: atomesh, version: "04b120d3040e0dd1a6915e88a3a3c1f588e6684a" }
-  kv-p2p-transfer: mooncake
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 1P1D TP4
-      - conc-list: [ 1, 2, 4, 8, 16, 32, 64, 128, 256 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=0"
-
-# Original config key: minimaxm3-fp4-mi355x-atom-disagg
-minimaxm3-fp4-mi355x-atom-disagg-deprecated-1k1k:
-  image: rocm/atom-dev:nightly_202607011530
-  model: amd/MiniMax-M3-MXFP4
-  model-prefix: minimaxm3
-  runner: cluster:mi355x-amds
-  precision: fp4
-  framework: atom-disagg
-  router: { name: atomesh, version: "04b120d3040e0dd1a6915e88a3a3c1f588e6684a" }
-  kv-p2p-transfer: mooncake
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 1P1D TP4
-      - conc-list: [ 1, 2, 4, 8, 16, 32, 64, 128, 256 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=0"
-
-# MiniMax-M3 MXFP8 MI300X day-zero recipe. Reuse the dedicated ROCm image and
-# MI355X serving shape, but retain the default BF16 KV cache because this
-# checkpoint lacks calibrated ROCm FP8 attention scales. TP8-only, plain
-# (non-expert-parallel) search space across the full conc range: EP8
-# (--enable-expert-parallel) produces garbage/incoherent output on this
-# MXFP8+gfx942 combination (confirmed locally: TP8/EP8 returns garbled tokens
-# even on trivial prompts, TP8/EP1 answers correctly), and was already the
-# lower-throughput topology where measured.
-# Original config key: minimaxm3-fp8-mi300x-vllm
-minimaxm3-fp8-mi300x-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:nightly-4559c43a9526597c00cbcc4f59979496500268d1
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: mi300x
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 256 }
-
-# Original config key: minimaxm3-fp8-mi300x-vllm-mtp
-minimaxm3-fp8-mi300x-vllm-mtp-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:nightly-b53b1c7ffe7aebdafd0876350f30e51d1226c92a
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: mi300x
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 128, spec-decoding: mtp }
-      - { tp: 8, ep: 8, conc-start: 256, conc-end: 256, spec-decoding: mtp }
-
-# Original config key: minimaxm3-fp8-mi325x-vllm
-minimaxm3-fp8-mi325x-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:minimax-m3
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: mi325x
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 64 }
-      - { tp: 4, ep: 4, conc-start: 128, conc-end: 256 }
-      - { tp: 8, conc-start: 1, conc-end: 128 }
-      - { tp: 8, ep: 8, conc-start: 256, conc-end: 512 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 1024 }
-
-# Original config key: minimaxm3-fp8-mi325x-vllm-mtp
-minimaxm3-fp8-mi325x-vllm-mtp-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:minimax-m3
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: mi325x
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 4, ep: 4, conc-start: 128, conc-end: 256, spec-decoding: mtp }
-      - { tp: 8, conc-start: 1, conc-end: 128, spec-decoding: mtp }
-      - { tp: 8, ep: 8, conc-start: 256, conc-end: 256, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 512, spec-decoding: mtp }
-
-# Original config key: minimaxm3-fp8-mi355x-vllm-disagg
-minimaxm3-fp8-mi355x-vllm-disagg-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:nightly-2afa3f7e950264bb179d030c23a1ed1f46558fd9
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: cluster:mi355x-amds
-  precision: fp8
-  framework: vllm-disagg
-  router: { name: vllm-router, version: "0.1.14" }
-  kv-p2p-transfer: moriio
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # Balanced 1P TP4 + 1D TP4 (2 nodes) across the full curve.
-      - spec-decoding: "none"
-        conc-list: [ 1, 2, 4, 8, 16, 32, 64, 128, 256, 512, 1024 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-      # Decode-heavy 1P TP4 + 2D TP4 (3 nodes): double the decode engines to
-      # absorb the decode-bound 1k1k tail at high concurrency.
-      - spec-decoding: "none"
-        conc-list: [ 256, 512, 1024, 2048 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-    # 8k1k is prefill-bound (8192-token prompts vs 1024 decode steps): pair the
-    # balanced layout with a prefill-heavy 2P/1D layout. Concurrency is capped at
-    # 512 so the multi-node eval policy (8k1k + conc >= 16, highest eligible conc)
-    # marks lm-eval at conc 512 — matching the range NVIDIA's aggregated 8k1k
-    # sweep tops out at and keeping the lm-eval async client stable.
-
-# Original config key: dsv4-fp8-mi325x-vllm
-dsv4-fp8-mi325x-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:nightly-09663abde0f50944a8d5ea30120666024b503faa
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  # mi325x (mi325x-amds pool) runners are down; use the mi325x-tw cluster.
-  runner: cluster:mi325x-tw
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 512 }
-
-# Original config key: dsv4-fp8-mi325x-vllm-mtp
-dsv4-fp8-mi325x-vllm-mtp-deprecated-1k1k:
-  image: vllm/vllm-openai-rocm:nightly-09663abde0f50944a8d5ea30120666024b503faa
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  # mi325x (mi325x-amds pool) runners are down; use the mi325x-tw cluster.
-  runner: cluster:mi325x-tw
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 512, spec-decoding: mtp }
-
-# Archive source: amd-dsv4-8k1k-master.yaml
-# Deprecated DeepSeek-V4-Pro 1.6T Single-turn 8k1k entries archived from
-# amd-master.yaml. Deprecated after Tuesday, September 8, 2026 (see the
-# Deprecation Notice in MODELS.md). Agentic coding is unaffected and stays
-# active for dsv4, including its MTP and DSpark arms; the model is not retired.
-# Removed from the active master config so sweep generation no longer selects them.
-
-# DeepSeek-V4-Pro FP8 single-node on MI300X (gfx942) via vLLM.
-# EXTRAPOLATED bring-up. sglang was abandoned: on gfx942 (no native FP4) the
-# dsv4 sglang backend's nvfp4 MoE / TileLang-MLA kernels have no gfx942 build
-# (gfx950/MI355X only). vLLM runs the checkpoint in FP8 via --quantization
-# deepseek_v4_fp8 (dequant FP4 MoE -> FP8), the same path the H200 dsv4 vLLM
-# recipe uses. Config mirrors the same-model dsv4-fp4-mi355x-vllm (TP8, conc
-# 4-512); the FP4->FP8 dequant (~1.05TB) fits 8x192GB only at TP8. Launch
-# script dsv4_fp8_mi300x.sh carries the deepseek_v4 + gfx942 AITER flags.
-dsv4-fp8-mi300x-vllm:
-  image: vllm/vllm-openai-rocm:nightly-09663abde0f50944a8d5ea30120666024b503faa
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: mi300x
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 8k1k KV fits only ~20x concurrency (9472-token requests), so conc512 is
-      # ~25x oversubscribed -> request timeouts. conc256 is proven green; cap here.
-      - { tp: 8, conc-start: 4, conc-end: 256 }
-
-# MTP variant of dsv4-fp8-mi300x-vllm. Mirrors the base recipe and adds
-# DeepSeek-V4 built-in MTP via --speculative-config (num_speculative_tokens=2),
-# routing to dsv4_fp8_mi300x_mtp.sh; benchmark uses --dsv4 chat-template
-# encoding (required for meaningful MTP acceptance).
-dsv4-fp8-mi300x-vllm-mtp:
-  image: vllm/vllm-openai-rocm:nightly-09663abde0f50944a8d5ea30120666024b503faa
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: mi300x
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 8k1k MTP KV is tighter than normal (draft model): conc256 failed here
-      # (19.2% req failures) and conc512 was 55.8%; conc128 passed cleanly.
-      # Cap 8k1k MTP at 128 (normal holds 256, 1k1k holds 512).
-      - { tp: 8, conc-start: 4, conc-end: 128, spec-decoding: mtp }
-
-dsv4-fp4-mi355x-sglang-disagg:
-  image: lmsysorg/sglang-rocm:v0.5.14-rocm720-mi35x-20260701
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: cluster:mi355x-amds
-  precision: fp4
-  framework: sglang-disagg
-  router: { name: sglang-router, version: "0.3.2" }
-  kv-p2p-transfer: mori
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # non-MTP configurations
-      # 1P1D pure TP8  (mori KV transfer)
-      - spec-decoding: "none"
-        conc-list: [ 1, 2, 4, 8, 16, 32, 64, 128 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=0"
-      # 1P1D DEP8  (mori KV transfer + mori MoE a2a, dp-attention)
-      - spec-decoding: "none"
-        conc-list: [ 256 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=0"
-
-      - spec-decoding: "none"
-        conc-list: [ 512 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=0"
-
-# MTP variant of dsv4-fp4-mi355x-sglang-disagg (base run 29821778206). Mirrors
-# the base disagg search space and topologies exactly (same 20260701 image,
-# mori KV transfer + MoE a2a, sglang-router --pd-disaggregation) and enables
-# DeepSeek-V4 built-in MTP on the decode side via spec-decoding: mtp — the same
-# enablement path as dsr1-fp8-mi355x-sglang-disagg-mtp. DECODE_MTP_SIZE=1 (draft
-# depth 1) on every arm. Depth 2 was tried on the pure-TP8 (bottom-of-curve) arm,
-# but this 20260701 build's 2-step EAGLE verification is lossy: greedy gsm8k fell
-# to ~0.85 (below the 0.91 dsv4 floor), while depth 1 is lossless (~0.96, matching
-# the spec-none base). So all arms use depth 1. The dsv4 ROCm MTP
-# CUDA-graph fix (sgl-project/sglang#26383, merged 2026-05-27) is already
-# present in the pinned 20260701 nightly, so no image bump is needed.
-dsv4-fp4-mi355x-sglang-disagg-mtp:
-  image: lmsysorg/sglang-rocm:v0.5.15-rocm720-mi35x-20260713
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: cluster:mi355x-amds
-  precision: fp4
-  framework: sglang-disagg
-  router: { name: sglang-router, version: "0.3.2" }
-  kv-p2p-transfer: mori
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # MTP configurations
-      # 1P1D pure TP8  (mori KV transfer)
-      - spec-decoding: "mtp"
-        conc-list: [ 1, 2, 4, 8, 16, 32, 64, 128 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=2"
-      # 1P1D DEP8  (mori KV transfer + mori MoE a2a, dp-attention)
-      - spec-decoding: "mtp"
-        conc-list: [ 256 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=1"
-
-      - spec-decoding: "mtp"
-        conc-list: [ 512 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=1"
-
-dsv4-fp4-mi355x-sglang:
-  image: lmsysorg/sglang-rocm:v0.5.14-rocm720-mi35x-20260706
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: mi355x
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, dp-attn: true, conc-start: 64, conc-end: 2048 }
-      - { tp: 4, dp-attn: true, conc-start: 16, conc-end: 128 }
-      - { tp: 4, dp-attn: false, conc-start: 1, conc-end: 32 }
-
-# MTP variant of dsv4-fp4-mi355x-sglang. Mirrors the base search space and adds
-# spec-decoding: mtp, which routes to dsv4_fp4_mi355x_sglang_mtp.sh (EAGLE
-# speculative decoding), per sgl-project/sglang#26383 ([AMD][DSV4] DSV4 MTP
-# graph + sparse triton attn optimizations, merged to main 2026-05-27). That PR
-# fixes the ROCm HIP-radix MTP CUDA-graph bug (the false-EOS symptom in sgl
-# #20404) and validates GSM8K 0.950 with MTP on.
-dsv4-fp4-mi355x-sglang-mtp:
-  image: lmsysorg/sglang-rocm:v0.5.19-rocm720-mi35x-20260907
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: mi355x
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, dp-attn: true, conc-start: 64, conc-end: 2048, spec-decoding: mtp }
-      - { tp: 8, dp-attn: false, conc-list: [1, 2, 4, 16, 32], spec-decoding: mtp }
-
-# DSv4 on MI355X via vLLM, using the official vllm/vllm-openai-rocm
-# nightly image. DSv4 base ROCm support (vllm-project/vllm#40871) merged
-# on 2026-05-05, so any nightly built after that includes the
-# DeepseekV4ForCausalLM model class.
-#
-# IMPORTANT: pin to a digest-suffixed nightly tag rather than the
-# floating `:nightly`. launch_mi355x-amds.sh caches enroot squashfs
-# files keyed on the image string and short-circuits re-import if the
-# file already exists, so the floating tag silently keeps a stale build
-# even after Docker Hub updates `:nightly`.
-#
-# DeepSeek-V4-Pro is FP4+FP8 mixed (FP4 MoE expert weights, FP8 for the
-# rest); InferenceX classifies this as fp4 — same as the sister sglang
-# and atom DSv4 mi355x entries below. The serving recipe enables AITER,
-# shared-expert fusion, and INT4 quick-reduce explicitly; the pinned nightly
-# then auto-selects its ROCm DeepSeek-V4 mHC, sparse-indexer, attention, and
-# compressor-GEMM optimizations. Shared-expert fusion (FHMoE) needs
-# vllm-project/vllm#53161, first shipped in this nightly; on earlier builds
-# it self-disabled because this checkpoint path was not eligible.
-dsv4-fp4-mi355x-vllm:
-  image: vllm/vllm-openai-rocm:nightly-d9105ea8001e0a6d77a96327d17515bb5791fb36
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: mi355x
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 512 }
-
-# MTP variant of dsv4-fp4-mi355x-vllm. Mirrors the base recipe's search space
-# and adds spec-decoding: mtp, which routes to dsv4_fp4_mi355x_vllm_mtp.sh
-# (--speculative-config '{"method":"mtp","num_speculative_tokens":2}'), per
-# vllm-project/vllm#43385 (ROCm DeepSeek-V4 MTP, merged 2026-05-24, included in
-# v0.22.0). Full conc 4-512 range maps the complete crossover curve: MTP wins
-# at low batch (PR perf data: +75% @ conc1, +38% @ conc8) and falls behind STP
-# above ~conc32 (-37% @ conc32). Image reuses the base entry's pinned ROCm
-# nightly, which includes the MTP path and subsequent DSV4 kernel updates.
-dsv4-fp4-mi355x-vllm-mtp:
-  image: vllm/vllm-openai-rocm:nightly-d9105ea8001e0a6d77a96327d17515bb5791fb36
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: mi355x
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 512, spec-decoding: mtp }
-
-dsv4-fp4-mi355x-atom:
-  image: rocm/atom-dev:nightly_202607231538
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: mi355x
-  precision: fp4
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # conc<64: tp4 dp-off (best at low conc); conc>=64: dp-on + TBO.
-      # tp8 dp-off low-conc kept as redundancy to recheck the crossover on the new image.
-      # dp-on cells always set max_req=conc (recipe .sh) to avoid OOM from the large
-      # default max_num_seqs (dp-attn keeps a full KV pool per rank); TBO stays conc>=64.
-      - { tp: 4, ep: 1, dp-attn: false, conc-start: 1,  conc-end: 64 }
-      - { tp: 4, ep: 1, dp-attn: true,  conc-start: 16, conc-end: 128 }
-      - { tp: 8, ep: 1, dp-attn: false, conc-start: 1,  conc-end: 64 }
-      - { tp: 8, ep: 1, dp-attn: true,  conc-start: 64, conc-end: 2048 }
-
-dsv4-fp4-mi355x-atom-mtp:
-  image: rocm/atom-dev:nightly_202607231538
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: mi355x
-  precision: fp4
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # Aligns ATOM official DSv4 MTP (models.json): MTP3 dp-off at low-mid conc,
-      # DPA MTP3 (dp-attn) at mid-high conc. dp-on starts at c32 (dp-attn helps from c32).
-      # prefill-only TBO on dp-attn cells at conc>=256 (measured crossover: TBO hurts
-      # c64/c128 -10~14%, helps c256+ +8~14%; MTP removes low-conc latency so TBO there is
-      # pure overhead). MTP-safe — only decode-TBO (--enable-tbo all) drops spec metadata.
-      - { tp: 8, ep: 1, dp-attn: false, conc-start: 4,  conc-end: 256,  spec-decoding: mtp }
-      - { tp: 8, ep: 1, dp-attn: true,  conc-start: 32, conc-end: 1024, spec-decoding: mtp }
-      # tp4 removed: tp4+MTP+dp-on OOMs — 806G/4=201G/GPU weight leaves KV budget
-      # NEGATIVE (-10.98GB) even at gpu-mem-util 0.90 (run 30277221626). #2327 tp4
-      # works only because it's non-MTP (no draft KV layer). MTP stays tp8-only.
-
-# MiniMax-M3 MXFP8 MI355X recipe:
-# https://github.com/vllm-project/recipes/commit/2a3728ed9892debfd767a72a58ebc90b33f186e5
-# MXFP8 runs from TP=4 on gfx950; block size 128 is mandatory for MSA.
-dsv4-fp4-mi355x-atom-disagg:
-  image: rocm/atom-dev:nightly_202606101403
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: mi355x
-  precision: fp4
-  framework: atom-disagg
-  router: { name: atomesh, version: "087b82d9c1f630e79149ba37e6213257ec9a76f8" }
-  kv-p2p-transfer: mooncake
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 2P1D DPA+TP8
-      - conc-list: [ 256, 512, 768, 1024, 2048 ]
-        prefill:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "PREFILL_NODES=2"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=1"
-      # 1P1D TP8
-      - conc-list: [ 4, 8, 16, 32, 64, 128 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-      # 1P1D TP8
-
-# DeepSeek-V4-Pro FP8 single-node on MI325X (gfx942) via vLLM.
-# EXTRAPOLATED bring-up. Same rationale as dsv4-fp8-mi300x-vllm: sglang has no
-# gfx942 build of the dsv4 nvfp4 MoE / TileLang-MLA kernels, so vLLM runs the
-# checkpoint in FP8 via --quantization deepseek_v4_fp8 (dequant FP4 MoE -> FP8),
-# the H200 dsv4 vLLM path. Config mirrors the same-model dsv4-fp4-mi355x-vllm
-# (TP8, conc 4-512); 8x256GB (2TB) has ample headroom for the ~1.05TB FP8
-# footprint. Launch script dsv4_fp8_mi325x.sh carries the deepseek_v4 + gfx942
-# AITER flags.
-dsv4-fp8-mi325x-vllm:
-  image: vllm/vllm-openai-rocm:nightly-09663abde0f50944a8d5ea30120666024b503faa
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  # mi325x (mi325x-amds pool) runners are down; use the mi325x-tw cluster.
-  runner: cluster:mi325x-tw
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 8k1k KV fits only ~20x concurrency (9472-token requests), so conc512 is
-      # ~25x oversubscribed -> request timeouts. conc256 is proven green; cap here.
-      - { tp: 8, conc-start: 4, conc-end: 256 }
-
-# MTP variant of dsv4-fp8-mi325x-vllm. Mirrors the base recipe and adds
-# DeepSeek-V4 built-in MTP via --speculative-config (num_speculative_tokens=2),
-# routing to dsv4_fp8_mi325x_mtp.sh; benchmark uses --dsv4 chat-template
-# encoding (required for meaningful MTP acceptance).
-dsv4-fp8-mi325x-vllm-mtp:
-  image: vllm/vllm-openai-rocm:nightly-09663abde0f50944a8d5ea30120666024b503faa
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  # mi325x (mi325x-amds pool) runners are down; use the mi325x-tw cluster.
-  runner: cluster:mi325x-tw
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 8k1k MTP KV is tighter than normal (draft model): conc256 is borderline
-      # (MI300X 19.2% req failures, flaked pass->fail across runs) and conc512
-      # is 55.8%. conc128 passed cleanly on the memory-tightest SKU (MI300X);
-      # cap 8k1k MTP at 128 (normal holds 256, 1k1k holds 512).
-      - { tp: 8, conc-start: 4, conc-end: 128, spec-decoding: mtp }
-
-# Archive source: amd-glm5-glm5.1-master.yaml
-# Deprecated GLM-5 / GLM-5.1 entries archived from amd-master.yaml.
-# Removed from the active master config so sweep generation no longer selects them.
-
-glm5-fp8-mi355x-sglang:
-  image: lmsysorg/sglang-rocm:v0.5.12-rocm720-mi35x-20260517
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: mi355x
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 4, conc-end: 256 }
-
-
-glm5-fp8-mi355x-sglang-mtp:
-  image: lmsysorg/sglang-rocm:v0.5.12-rocm720-mi35x-20260517
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: mi355x
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 4, conc-end: 128, spec-decoding: mtp }
-      - { tp: 8, conc-start: 4, conc-end: 8, spec-decoding: mtp }
-
-
-glm5-fp8-mi355x-sglang-disagg:
-  image: lmsysorg/sglang-rocm:v0.5.12.post1-rocm720-mi35x-20260523
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: cluster:mi355x-amds
-  precision: fp8
-  framework: sglang-disagg
-  router: { name: sglang-router, version: "0.3.2" }
-  kv-p2p-transfer: mori
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 1P+1D TP8/EP1 CI smoke sweep; dp-attn false (NSA / MoRI path)
-      - spec-decoding: "none"
-        conc-list: [ 8, 16, 32, 64, 128, 256, 512 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=0"
-
-
-glm5-fp8-mi355x-atom:
-  image: rocm/atom:rocm7.2.2_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom0.1.2.post
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: mi355x
-  precision: fp8
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 4, conc-end: 256 }
-      - { tp: 8, conc-start: 4, conc-end: 256 }
-
-
-glm5.1-fp4-mi355x-sglang:
-  image: lmsysorg/sglang-rocm:v0.5.12.post1-rocm720-mi35x-20260529
-  model: amd/GLM-5.1-MXFP4
-  model-prefix: glm5.1
-  runner: mi355x
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 2, conc-start: 4, conc-end: 256 }
-      - { tp: 4, conc-start: 4, conc-end: 16 }
-
-
-glm5.1-fp4-mi355x-atom:
-  image: rocm/atom:rocm7.2.2_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom0.1.2.post
-  model: amd/GLM-5.1-MXFP4
-  model-prefix: glm5.1
-  runner: mi355x
-  precision: fp4
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 4, conc-end: 256 }
-
-
-glm5-fp8-mi325x-sglang:
-  image: lmsysorg/sglang:v0.5.12-rocm720-mi30x
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: mi325x
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-
-glm5-fp8-mi325x-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.12-rocm720-mi30x
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: mi325x
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-
-# ============================================================================
-# Net-new agentic recipes from chore/agentx-v0.3 (no overlap with main entries).
-# Recipes that ALREADY existed on main were intentionally left at main's version
-# to preserve main behavior; PR-branch modifications to those recipes are NOT
-# brought in here.
-# ============================================================================
-
-# Archive source: amd-gptoss-master.yaml
-# Deprecated gpt-oss-120b entries archived from amd-master.yaml.
-# Removed from the active master config so sweep generation no longer selects them.
-
-gptoss-fp4-mi300x-vllm:
-  image: vllm/vllm-openai-rocm:v0.17.0
-  model: openai/gpt-oss-120b
-  model-prefix: gptoss
-  runner: mi300x
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 64, conc-end: 256 }
-      - { tp: 2, conc-start: 4, conc-end: 64 }
-      - { tp: 4, conc-start: 4, conc-end: 64 }
-      - { tp: 8, conc-start: 1, conc-end: 16 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 4, conc-end: 64 }
-      - { tp: 2, conc-start: 4, conc-end: 64 }
-      - { tp: 4, conc-start: 4, conc-end: 64 }
-      - { tp: 8, conc-start: 1, conc-end: 16 }
-
-gptoss-fp4-mi325x-vllm:
-  image: vllm/vllm-openai-rocm:v0.22.0
-  model: openai/gpt-oss-120b
-  model-prefix: gptoss
-  runner: mi325x
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 4, conc-end: 64 }
-      - { tp: 2, conc-start: 4, conc-end: 64 }
-      - { tp: 4, conc-start: 4, conc-end: 64 }
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 4, conc-end: 64 }
-      - { tp: 2, conc-start: 4, conc-end: 8 }
-      - { tp: 4, conc-start: 4, conc-end: 8 }
-      - { tp: 8, conc-start: 4, conc-end: 16 }
-
-gptoss-fp4-mi355x-vllm:
-  image: vllm/vllm-openai-rocm:v0.22.0
-  model: openai/gpt-oss-120b
-  model-prefix: gptoss
-  runner: mi355x
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 4, conc-end: 128 }
-      - { tp: 4, conc-start: 4, conc-end: 8 }
-      - { tp: 8, conc-start: 4, conc-end: 16 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 4, conc-end: 128 }
-      - { tp: 4, conc-start: 4, conc-end: 4 }
-      - { tp: 8, conc-start: 4, conc-end: 8 }
-
-gptoss-fp4-mi355x-atom:
-  image: rocm/atom:rocm7.2.2_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom0.1.2.post
-  model: openai/gpt-oss-120b
-  model-prefix: gptoss
-  runner: mi355x
-  precision: fp4
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 16, conc-end: 256 }
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 32 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 4, conc-end: 256 }
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 16 }
-
-# Archive source: amd-kimik2.5-8k1k-master.yaml
-# Deprecated Kimi-K2.5/2.6/2.7-Code Single-turn 8k1k entries archived from
-# amd-master.yaml. Deprecated after Thursday, August 6, 2026 (see the
-# Deprecation Notice in MODELS.md). Combined with the agentic-coding
-# deprecation of August 3, 2026, this leaves kimik2.5 with no active scenario:
-# the model is fully retired after August 6, 2026.
-# Removed from the active master config so sweep generation no longer selects them.
-
-kimik2.5-int4-mi355x-vllm:
-  image: vllm/vllm-openai-rocm:v0.24.0
-  model: moonshotai/Kimi-K2.5
-  model-prefix: kimik2.5
-  runner: mi355x
-  precision: int4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 128 }
-      - { tp: 4, conc-start: 4, conc-end: 128 }
-
-kimik2.5-int4-mi325x-vllm:
-  image: vllm/vllm-openai-rocm:v0.21.0
-  model: moonshotai/Kimi-K2.5
-  model-prefix: kimik2.5
-  runner: mi325x
-  precision: int4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-kimik2.5-int4-mi300x-vllm:
-  image: vllm/vllm-openai-rocm:v0.21.0
-  model: moonshotai/Kimi-K2.5
-  model-prefix: kimik2.5
-  runner: mi300x
-  precision: int4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-kimik2.5-fp4-mi355x-vllm:
-  image: vllm/vllm-openai-rocm:nightly-387189c42997b27e2c04b5d97ef8190ffa2bf909
-  model: amd/Kimi-K2.5-MXFP4
-  model-prefix: kimik2.5
-  runner: mi355x
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 128 }
-      - { tp: 4, conc-start: 4, conc-end: 128 }
-
-kimik2.5-fp4-mi355x-atom:
-  image: rocm/atom:rocm7.2.4_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom0.1.4_202607091539
-  model: amd/Kimi-K2.5-MXFP4
-  model-prefix: kimik2.5
-  runner: mi355x
-  precision: fp4
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 4, conc-end: 128 }
-
-kimik2.5-fp4-mi355x-atom-disagg:
-  image: rocm/atom:rocm7.2.4_ubuntu24.04_py3.12_pytorch_release_2.10.0_atomesh_202607121715
-  model: amd/Kimi-K2.5-MXFP4
-  model-prefix: kimik2.5
-  runner: cluster:mi355x-amds
-  precision: fp4
-  framework: atom-disagg
-  kv-p2p-transfer: mooncake
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-      # 1P(tp4)1D(tp4)
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - spec-decoding: "none"
-        conc-list: [ 16, 32, 64, 128, 256 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-      # 1P(tp4)2D(tp8)
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - spec-decoding: "none"
-        conc-list: [ 1, 2, 4, 8, 16 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-
-kimik2.5-fp4-mi355x-vllm-disagg:
-  image: vllm/vllm-openai-rocm:nightly-2afa3f7e950264bb179d030c23a1ed1f46558fd9
-  model: amd/Kimi-K2.5-MXFP4
-  model-prefix: kimik2.5
-  runner: cluster:mi355x-amds
-  precision: fp4
-  framework: vllm-disagg
-  router: { name: vllm-router, version: "0.1.14" }
-  kv-p2p-transfer: moriio
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    # All workers TP4 (real-weight sweep: TP8 decode is no better than TP4).
-    # Split across P/D topologies: 1D is stable at low conc; the high-conc tail
-    # runs on 2D so decode load is spread across two engines.
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 1P(TP4) 1D(TP4) = 2 nodes. Low concurrency only.
-      - spec-decoding: "none"
-        conc-list: [ 1, 2, 4 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-      # 1P(TP4) 2D(TP4) = 3 nodes. High concurrency (decode KV/load spread over 2D).
-      - spec-decoding: "none"
-        conc-list: [ 8, 16, 32, 64, 128, 256, 512 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-
-# Archive source: amd-minimaxm2.5-m2.7-master.yaml
-# Deprecated MiniMax M2.5/M2.7 entries archived from amd-master.yaml.
-# Removed from the active master config so sweep generation no longer selects them.
-
-
-minimaxm2.5-fp8-mi355x-vllm:
-  image: vllm/vllm-openai-rocm:v0.22.0
-  model: MiniMaxAI/MiniMax-M2.5
-  model-prefix: minimaxm2.5
-  runner: mi355x
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 2, ep: 2, conc-start: 2, conc-end: 512 }
-      - { tp: 4, ep: 4, conc-start: 4, conc-end: 256 }
-      - { tp: 8, ep: 8, conc-start: 2, conc-end: 2 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 2, ep: 2, conc-start: 2, conc-end: 256 }
-      - { tp: 4, ep: 4, conc-start: 4, conc-end: 512 }
-      - { tp: 8, ep: 8, conc-start: 2, conc-end: 2 }
-
-# Diverged from minimaxm2.5-fp8-mi355x-vllm (agentic-coding sibling). Reasons below;
-# the original minimaxm2.5-fp8-mi355x-vllm entry is left identical to origin/main so
-# its fixed-seq-len sweep is unaffected.
-#   - image: 'vllm/vllm-openai-rocm:v0.19.0' -> 'vllm/vllm-openai-rocm:nightly-51f22dcfd068fe8f1e3192da2a1e825b930223cf'
-minimaxm2.5-fp8-mi355x-atom:
-  image: rocm/atom:rocm7.2.2_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom0.1.2.post
-  model: MiniMaxAI/MiniMax-M2.5
-  model-prefix: minimaxm2.5
-  runner: mi355x
-  precision: fp8
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 2, conc-start: 4, conc-end: 256 }
-      - { tp: 4, conc-start: 4, conc-end: 256 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 2, conc-start: 4, conc-end: 256 }
-      - { tp: 4, conc-start: 4, conc-end: 256 }
-
-minimaxm2.5-fp4-mi355x-atom:
-  image: rocm/atom:rocm7.2.2_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom0.1.2.post
-  model: amd/MiniMax-M2.5-MXFP4
-  model-prefix: minimaxm2.5
-  runner: mi355x
-  precision: fp4
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 4, conc-end: 1024 }
-      - { tp: 2, conc-start: 4, conc-end: 1024 }
-      - { tp: 4, conc-start: 4, conc-end: 128 }
-      - { tp: 8, conc-start: 4, conc-end: 16 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 4, conc-end: 1024 }
-      - { tp: 2, conc-start: 4, conc-end: 1024 }
-      - { tp: 4, conc-start: 4, conc-end: 128 }
-      - { tp: 8, conc-start: 4, conc-end: 16 }
-
-minimaxm2.5-fp4-mi355x-vllm:
-  image: vllm/vllm-openai-rocm:v0.22.0
-  model: amd/MiniMax-M2.5-MXFP4
-  model-prefix: minimaxm2.5
-  runner: mi355x
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 4, conc-end: 32 }
-      - { tp: 2, conc-start: 4, conc-end: 64 }
-      - { tp: 4, conc-start: 4, conc-end: 64 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 4, conc-end: 32 }
-      - { tp: 2, conc-start: 4, conc-end: 64 }
-      - { tp: 4, conc-start: 4, conc-end: 64 }
-
-minimaxm2.5-fp8-mi300x-vllm:
-  image: vllm/vllm-openai-rocm:v0.21.0
-  model: MiniMaxAI/MiniMax-M2.5
-  model-prefix: minimaxm2.5
-  runner: mi300x
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 2, conc-start: 4, conc-end: 64 }
-      - { tp: 4, conc-start: 4, conc-end: 64 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 2, conc-start: 4, conc-end: 64 }
-      - { tp: 4, conc-start: 4, conc-end: 64 }
-
-# Diverged from minimaxm2.5-fp8-mi300x-vllm (agentic-coding sibling). Reasons below;
-# the original minimaxm2.5-fp8-mi300x-vllm entry is left identical to origin/main so
-# its fixed-seq-len sweep is unaffected.
-#   - image: 'vllm/vllm-openai-rocm:v0.16.0' -> 'vllm/vllm-openai-rocm:nightly-51f22dcfd068fe8f1e3192da2a1e825b930223cf'
-minimaxm2.5-fp8-mi325x-vllm:
-  image: vllm/vllm-openai-rocm:v0.22.0
-  model: MiniMaxAI/MiniMax-M2.5
-  model-prefix: minimaxm2.5
-  runner: mi325x
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 2, conc-start: 4, conc-end: 64 }
-      - { tp: 8, ep: 8, conc-start: 4, conc-end: 512 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 2, conc-start: 4, conc-end: 64 }
-      - { tp: 8, ep: 8, conc-start: 4, conc-end: 256 }
-
-# Diverged from minimaxm2.5-fp8-mi325x-vllm (agentic-coding sibling). Reasons below;
-# the original minimaxm2.5-fp8-mi325x-vllm entry is left identical to origin/main so
-# its fixed-seq-len sweep is unaffected.
-#   - image: 'vllm/vllm-openai-rocm:v0.18.0' -> 'vllm/vllm-openai-rocm:nightly-51f22dcfd068fe8f1e3192da2a1e825b930223cf'
-minimaxm2.5-fp8-mi355x-vllm-disagg:
-  image: vllm/vllm-openai-rocm:nightly-a6682d1d259cca69a9ae737ea5608fbbe7520031
-  model: MiniMaxAI/MiniMax-M2.5
-  model-prefix: minimaxm2.5
-  runner: cluster:mi355x-amds
-  precision: fp8
-  framework: vllm-disagg
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 1P2D: 1 prefill node (co-located with proxy) + 2 decode nodes = 3 nodes total
-      # Prefill also needs EP=8: MiniMax M2.5 expert intermediate_size=1536,
-      # TP8 shards to 192 which is not divisible by FP8 block_n=128.
-      - spec-decoding: "none"
-        conc-list: [ 8, 16, 32, 64, 128, 256, 512 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-          - "VLLM_MORIIO_CONNECTOR_READ_MODE=1"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - spec-decoding: "none"
-        conc-list: [ 8, 16, 32, 64, 128, 256, 512 ]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-          - "VLLM_MORIIO_CONNECTOR_READ_MODE=1"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-
-# Archive source: amd-minimaxm3-8k1k-master.yaml
-# Deprecated MiniMax-M3 Single-turn 8k1k entries archived from amd-master.yaml.
-# The scenario was deprecated after Monday, August 3, 2026 (see the Deprecation
-# Notice in MODELS.md). Agentic coding remains active for this model.
-# Removed from the active master config so sweep generation no longer selects them.
-
-minimaxm3-fp8-mi355x-vllm:
-  image: vllm/vllm-openai-rocm:nightly-9e57de7197f234f9d9187715d96e07e007048c0f
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: mi355x
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 512 }
-
-# EAGLE3 speculative-decoding (spec-decoding: mtp) variant of
-# minimaxm3-fp8-mi355x-vllm, pairing MiniMaxAI/MiniMax-M3-MXFP8 with the
-# Inferact/MiniMax-M3-EAGLE3 draft head (3 speculative tokens). No
-# attention_backend override is needed — the server runs on TRITON_ATTN, so
-# the FlashInfer page-128/MHA limitation that forced FLASH_ATTN on Blackwell
-# does not apply here. Search space mirrors the non-MTP entry trimmed at the
-# extreme-concurrency end, identical to the minimaxm3-fp8-b300-vllm-mtp /
-# b200-vllm-mtp precedent: spec decode pays off at low/mid concurrency while
-# acceptance dilutes in big batches, and the draft weights + draft KV shave
-# headroom — tp2-ep2 is dropped since its KV headroom was already thin.
-minimaxm3-fp8-mi355x-vllm-mtp:
-  image: vllm/vllm-openai-rocm:nightly-09663abde0f50944a8d5ea30120666024b503faa
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: mi355x
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 512, spec-decoding: mtp }
-
-# MiniMax-M3 MXFP4 MI355X vLLM disaggregated (prefill/decode) config.
-minimaxm3-fp4-mi355x-vllm-disagg:
-  image: vllm/vllm-openai-rocm:nightly-2dfaae752b4db0d43cfc0715c780e33be030d0f1
-  model: amd/MiniMax-M3-MXFP4
-  model-prefix: minimaxm3
-  runner: cluster:mi355x-amds
-  precision: fp4
-  framework: vllm-disagg
-  router: { name: vllm-router, version: "0.1.14" }
-  kv-p2p-transfer: moriio
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 1P TP4 + 1D TP4 (2 nodes total), conc sweep 1..256 (single job, looped)
-      - spec-decoding: "none"
-        conc-list: [ 1, 2, 4, 8, 16, 32, 64, 128, 256 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-
-# MiniMax-M3 MXFP4 MI355X vLLM recipe. The pinned nightly includes upstream
-# MiniMax-M3 Quark MXFP4 support (vllm-project/vllm#45794). Use the text-only
-# language-model path and mirror the MXFP8 MI355X search space for a direct
-# precision comparison.
-minimaxm3-fp4-mi355x-vllm:
-  image: vllm/vllm-openai-rocm:nightly-69715823df89b11ee684b84066390cbb9092d5c1
-  model: amd/MiniMax-M3-MXFP4
-  model-prefix: minimaxm3
-  runner: mi355x
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 2, conc-start: 1, conc-end: 512 }
-      - { tp: 4, conc-start: 1, conc-end: 64 }
-
-# EAGLE3 speculative-decoding variant of minimaxm3-fp4-mi355x-vllm. Pair the
-# amd/MiniMax-M3-MXFP4 target with Inferact/MiniMax-M3-EAGLE3 and three draft
-# tokens. Search space mirrors the MI355X MXFP8 MTP entry, trimming the base
-# FP4 sweep at extreme concurrency where speculative decoding loses value.
-minimaxm3-fp4-mi355x-vllm-mtp:
-  image: vllm/vllm-openai-rocm:nightly-3f5a1e1733200760169ff31ebe60a271072b199e
-  model: amd/MiniMax-M3-MXFP4
-  model-prefix: minimaxm3
-  runner: mi355x
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 8, ep: 8, conc-start: 1, conc-end: 256, spec-decoding: mtp }
-      - { tp: 4, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 128, conc-end: 256, spec-decoding: mtp }
-
-# MiniMax-M3 MXFP4 MI355X atom recipe:
-# https://github.com/ROCm/ATOM/blob/5d42d49f9e4292e5b61475917e92e7ec1b1dacb7/recipes/MiniMax-M3.md
-# block size 128 is mandatory for MSA. TP4 on a single gfx950 node, per the recipe.
-minimaxm3-fp4-mi355x-atom:
-  image: rocm/atom-dev:nightly_202607011530
-  model: amd/MiniMax-M3-MXFP4
-  model-prefix: minimaxm3
-  runner: mi355x
-  precision: fp4
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 256 }
-
-minimaxm3-fp4-mi355x-atom-mtp:
-  image: rocm/atom-dev:nightly_202607011530
-  model: amd/MiniMax-M3-MXFP4
-  model-prefix: minimaxm3
-  runner: mi355x
-  precision: fp4
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 256, spec-decoding: mtp  }
-
-minimaxm3-fp8-mi355x-atom:
-  image: rocm/atom-dev:nightly_202607011530
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: mi355x
-  precision: fp8
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 256 }
-
-minimaxm3-fp8-mi355x-atom-mtp:
-  image: rocm/atom-dev:nightly_202607011530
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: mi355x
-  precision: fp8
-  framework: atom
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 256, spec-decoding: mtp }
-
-minimaxm3-fp8-mi355x-atom-disagg:
-  image: rocm/atom-dev:nightly_202607011530
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: cluster:mi355x-amds
-  precision: fp8
-  framework: atom-disagg
-  router: { name: atomesh, version: "04b120d3040e0dd1a6915e88a3a3c1f588e6684a" }
-  kv-p2p-transfer: mooncake
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 1P1D TP4
-      - conc-list: [ 1, 2, 4, 8, 16, 32, 64, 128, 256 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=0"
-      # 2P1D, DPA TP4
-      - conc-list: [ 256, 512, 768, 1024 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "PREFILL_NODES=2"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=0"
-      # 1P1D TP4
-
-minimaxm3-fp4-mi355x-atom-disagg:
-  image: rocm/atom-dev:nightly_202607011530
-  model: amd/MiniMax-M3-MXFP4
-  model-prefix: minimaxm3
-  runner: cluster:mi355x-amds
-  precision: fp4
-  framework: atom-disagg
-  router: { name: atomesh, version: "04b120d3040e0dd1a6915e88a3a3c1f588e6684a" }
-  kv-p2p-transfer: mooncake
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 1P1D TP4
-      - conc-list: [ 1, 2, 4, 8, 16, 32, 64, 128, 256 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=0"
-      # 2P1D, DPA TP4
-      - conc-list: [ 256, 512, 768, 1024 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "PREFILL_NODES=2"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MTP_SIZE=0"
-      # 1P1D TP4
-
-minimaxm3-fp8-mi300x-vllm:
-  image: vllm/vllm-openai-rocm:nightly-4559c43a9526597c00cbcc4f59979496500268d1
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: mi300x
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 256 }
-
-# EAGLE3 speculative-decoding (spec-decoding: mtp) variant of
-# minimaxm3-fp8-mi300x-vllm, pairing MiniMaxAI/MiniMax-M3-MXFP8 with the
-# Inferact/MiniMax-M3-EAGLE3 draft head (3 speculative tokens). Same TP8-only
-# search space as the non-MTP MI300X entry (gfx942 192 GB is memory-tight, like
-# H100), with the TP8 latency rows started at conc 1 to capture single-request
-# latency — matching the H100/MI355X MTP recipes. The pinned ROCm nightly
-# includes upstream SupportsEagle3 support for the AMD MiniMax-M3 model.
-minimaxm3-fp8-mi300x-vllm-mtp:
-  image: vllm/vllm-openai-rocm:nightly-b53b1c7ffe7aebdafd0876350f30e51d1226c92a
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: mi300x
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 8, ep: 8, conc-start: 128, conc-end: 256, spec-decoding: mtp }
-
-# MiniMax-M3 MXFP8 MI325X day-zero recipe. Reuse the dedicated ROCm image
-# and serving flags validated on MI355X, with the H200 search space: TP4 and
-# TP8 latency, TP4/TP8 expert parallelism, and TP8 data-parallel attention.
-minimaxm3-fp8-mi325x-vllm:
-  image: vllm/vllm-openai-rocm:minimax-m3
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: mi325x
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 32 }
-      - { tp: 8, conc-start: 1, conc-end: 128 }
-      - { tp: 8, ep: 8, conc-start: 256, conc-end: 256 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 512 }
-
-# EAGLE3 speculative-decoding (spec-decoding: mtp) variant of
-# minimaxm3-fp8-mi325x-vllm, pairing MiniMaxAI/MiniMax-M3-MXFP8 with the
-# Inferact/MiniMax-M3-EAGLE3 draft head (3 speculative tokens). Same H200-style
-# search space as the non-MTP MI325X entry, trimmed at the extreme-concurrency
-# end with TP-only latency rows started at conc 1 (matching the H200/MI355X MTP
-# recipes). Runs with CUDA graphs (no --enforce-eager, VLLM_USE_BREAKABLE_CUDAGRAPH=0,
-# BF16 KV on gfx942). The shipped ROCm image lacks SupportsEagle3 on the AMD
-# MiniMax-M3 model, so the recipe applies that fix in-place at runtime
-# (functionstackx/vllm#1, upstream vllm-project/vllm#45546; validated green on
-# MI355X/MI300X) before serving.
-minimaxm3-fp8-mi325x-vllm-mtp:
-  image: vllm/vllm-openai-rocm:minimax-m3
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: mi325x
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 32, spec-decoding: mtp }
-      - { tp: 8, conc-start: 1, conc-end: 128, spec-decoding: mtp }
-      - { tp: 8, ep: 8, conc-start: 256, conc-end: 256, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 256, spec-decoding: mtp }
-
-# MiniMax-M3 MXFP8 MI355X vLLM disaggregated (prefill/decode) sweep on the
-# day-zero ROCm image, over the MoRI-IO KV-transfer pipeline (MoRI-patch-removal
-# infra #1585). All workers are TP4, no EP: the single-node M3 MXFP8 recipe
-# (minimaxm3-fp8-mi355x-vllm, PR #2003) found plain TP4 beats both TP8 and
-# TP4/EP4 on tok/s/GPU for this model on gfx950, so prefill and decode both use
-# TP4 and we tune the prefill:decode worker ratio (xP:yD) instead of TP. The
-# mi355x-disagg pool has 3 nodes and the launcher places one worker per node
-# (NUM_NODES = xP + yD), so every layout keeps xP + yD <= 3:
-#   - 1P-TP4 / 1D-TP4  (2 nodes): balanced, full concurrency curve.
-#   - 1P-TP4 / 2D-TP4  (3 nodes): decode-heavy, for the decode-bound 1k1k tail.
-#   - 2P-TP4 / 1D-TP4  (3 nodes): prefill-heavy, for the prefill-bound 8k1k tail.
-# Per-worker serve flags live in
-# benchmarks/multi_node/amd_utils/models_vllm.yaml (MiniMax-M3-MXFP8).
-minimaxm3-fp8-mi355x-vllm-disagg:
-  image: vllm/vllm-openai-rocm:nightly-2afa3f7e950264bb179d030c23a1ed1f46558fd9
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: cluster:mi355x-amds
-  precision: fp8
-  framework: vllm-disagg
-  router: { name: vllm-router, version: "0.1.14" }
-  kv-p2p-transfer: moriio
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # Balanced 1P TP4 + 1D TP4 (2 nodes) across the full curve.
-      - spec-decoding: "none"
-        conc-list: [ 1, 2, 4, 8, 16, 32, 64, 128, 256, 512 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-      # Prefill-heavy 2P TP4 + 1D TP4 (3 nodes): two half-node TP4 prefill workers
-      # keep the single TP4 decode engine fed for the prefill-bound 8k1k tail.
-      - spec-decoding: "none"
-        conc-list: [ 128, 256, 512 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=2"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-
-# Archive source: amd-qwen3.5-bf16-master.yaml
-# Deprecated Qwen3.5-397B-A17B bf16 entries archived from amd-master.yaml.
-# All bf16 recipes, in every scenario, were deprecated after Monday,
-# August 3, 2026 (see the Deprecation Notice in MODELS.md). fp8 and fp4
-# recipes remain active.
-# Removed from the active master config so sweep generation no longer selects them.
-
-
-qwen3.5-bf16-mi355x-sglang:
-  image: lmsysorg/sglang-rocm:v0.5.12-rocm720-mi35x-20260517
-  model: Qwen/Qwen3.5-397B-A17B
-  model-prefix: qwen3.5
-  runner: mi355x
-  precision: bf16
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 256 }
-
-qwen3.5-bf16-mi355x-sglang-mtp:
-  image: lmsysorg/sglang-rocm:v0.5.12-rocm720-mi35x-20260517
-  model: Qwen/Qwen3.5-397B-A17B
-  model-prefix: qwen3.5
-  runner: mi355x
-  precision: bf16
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 256, spec-decoding: mtp }
-
-qwen3.5-bf16-mi300x-sglang:
-  image: lmsysorg/sglang:v0.5.12-rocm720-mi30x
-  model: Qwen/Qwen3.5-397B-A17B
-  model-prefix: qwen3.5
-  runner: mi300x
-  precision: bf16
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-qwen3.5-bf16-mi325x-sglang:
-  image: lmsysorg/sglang:v0.5.12-rocm720-mi30x
-  model: Qwen/Qwen3.5-397B-A17B
-  model-prefix: qwen3.5
-  runner: mi325x
-  precision: bf16
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-qwen3.5-bf16-mi325x-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.12-rocm720-mi30x
-  model: Qwen/Qwen3.5-397B-A17B
-  model-prefix: qwen3.5
-  runner: mi325x
-  precision: bf16
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64, spec-decoding: mtp }
diff --git a/configs/deprecated/nvidia-master.yaml b/configs/deprecated/nvidia-master.yaml
deleted file mode 100644
index 9029f6235a..0000000000
--- a/configs/deprecated/nvidia-master.yaml
+++ /dev/null
@@ -1,16588 +0,0 @@
-# Deprecated NVIDIA configurations; historical reference only.
-# Active configs live in configs/nvidia-master.yaml.
-# Colliding historical 1k1k keys have a -deprecated-1k1k suffix;
-# their original keys are recorded in comments. Preserve each archived version.
-
-# Archive source: nvidia-1k1k-master.yaml
-# Deprecated 1k1k fixed-sequence entries archived from nvidia-master.yaml.
-# Removed from the active master config so sweep generation no longer selects them.
-
-dsr1-fp4-b200-dynamo-trt:
-  image: nvcr.io/nvidia/ai-dynamo/tensorrtllm-runtime:0.8.1.post1
-  model: deepseek-r1-fp4
-  model-prefix: dsr1
-  runner: cluster:b200-dgxc
-  precision: fp4
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "0.8.1.post1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - spec-decoding: "mtp"
-        conc-list: [1214]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp4/1k1k/mtp/ctx1_gen2_dep8_batch64_eplb0_mtp2.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp4/1k1k/mtp/ctx1_gen2_dep8_batch64_eplb0_mtp2.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [875]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp4/1k1k/mtp/ctx1_gen5_dep8_batch16_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp4/1k1k/mtp/ctx1_gen5_dep8_batch16_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [6]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp4/1k1k/mtp/ctx1_gen5_tep8_batch1_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp4/1k1k/mtp/ctx1_gen5_tep8_batch1_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [10, 15, 25, 45, 90, 180]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp4/1k1k/mtp/ctx1_gen5_tep8_batch32_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp4/1k1k/mtp/ctx1_gen5_tep8_batch32_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [ 4968 ]
-        prefill:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp4/1k1k/mtp/ctx3_gen4_dep8_batch128_eplb0_mtp1.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp4/1k1k/mtp/ctx3_gen4_dep8_batch128_eplb0_mtp1.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [10860]
-        prefill:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp4/1k1k/mtp/ctx3_gen5_dep4_batch512_eplb0_mtp1.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp4/1k1k/mtp/ctx3_gen5_dep4_batch512_eplb0_mtp1.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: true
-
-      # Non-MTP configurations
-      - conc-list: [4096]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp4/1k1k/stp/ctx1_gen1_dep8_batch512_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp4/1k1k/stp/ctx1_gen1_dep8_batch512_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [2192]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp4/1k1k/stp/ctx1_gen2_dep8_batch128_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp4/1k1k/stp/ctx1_gen2_dep8_batch128_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [1365]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp4/1k1k/stp/ctx1_gen5_dep8_batch32_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp4/1k1k/stp/ctx1_gen5_dep8_batch32_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [6]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp4/1k1k/stp/ctx1_gen5_tep8_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp4/1k1k/stp/ctx1_gen5_tep8_batch1_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [10, 15, 25, 45, 90, 180]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp4/1k1k/stp/ctx1_gen5_tep8_batch32_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp4/1k1k/stp/ctx1_gen5_tep8_batch32_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [450]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp4/1k1k/stp/ctx1_gen6_tep8_batch64_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp4/1k1k/stp/ctx1_gen6_tep8_batch64_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 6
-          tp: 8
-          ep: 8
-          dp-attn: false
-
-dsr1-fp8-b200-dynamo-trt:
-  image: nvcr.io/nvidia/ai-dynamo/tensorrtllm-runtime:0.8.1.post2
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: cluster:b200-dgxc
-  precision: fp8
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "0.8.1.post2" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # MTP configurations - Low latency (TP attention)
-      - spec-decoding: "mtp"
-        conc-list: [8]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp8/1k1k/mtp/ctx1_gen8_tp8_batch1_eplb0_mtp3_8.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp8/1k1k/mtp/ctx1_gen8_tp8_batch1_eplb0_mtp3_8.yaml"
-        decode:
-          num-worker: 8
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [32]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp8/1k1k/mtp/ctx1_gen8_tp8_batch4_eplb0_mtp3_32.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp8/1k1k/mtp/ctx1_gen8_tp8_batch4_eplb0_mtp3_32.yaml"
-        decode:
-          num-worker: 8
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [64]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp8/1k1k/mtp/ctx1_gen8_tp8_batch8_eplb0_mtp3_64.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp8/1k1k/mtp/ctx1_gen8_tp8_batch8_eplb0_mtp3_64.yaml"
-        decode:
-          num-worker: 8
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [256]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp8/1k1k/mtp/ctx1_gen8_tp8_batch32_eplb0_mtp3_256.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp8/1k1k/mtp/ctx1_gen8_tp8_batch32_eplb0_mtp3_256.yaml"
-        decode:
-          num-worker: 8
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # MTP configurations - High throughput (DP attention)
-      - spec-decoding: "mtp"
-        conc-list: [896]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp8/1k1k/mtp/ctx1_gen7_dep8_batch128_eplb0_mtp3_896.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp8/1k1k/mtp/ctx1_gen7_dep8_batch128_eplb0_mtp3_896.yaml"
-        decode:
-          num-worker: 7
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [1024]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp8/1k1k/mtp/ctx1_gen4_dep8_batch256_eplb0_mtp3_1024.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp8/1k1k/mtp/ctx1_gen4_dep8_batch256_eplb0_mtp3_1024.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [1184]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp8/1k1k/mtp/ctx1_gen3_dep8_batch384_eplb0_mtp3_1184.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp8/1k1k/mtp/ctx1_gen3_dep8_batch384_eplb0_mtp3_1184.yaml"
-        decode:
-          num-worker: 3
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [1600]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp8/1k1k/mtp/ctx1_gen2_dep8_batch768_eplb0_mtp2_1600.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp8/1k1k/mtp/ctx1_gen2_dep8_batch768_eplb0_mtp2_1600.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-      # Non-MTP (STP) configurations - Low latency (TP attention)
-      - conc-list: [4]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp8/1k1k/stp/ctx1_gen3_tp8_batch1024_eplb0_mtp0_4.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp8/1k1k/stp/ctx1_gen3_tp8_batch1024_eplb0_mtp0_4.yaml"
-        decode:
-          num-worker: 3
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [32]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp8/1k1k/stp/ctx1_gen3_tp8_batch1024_eplb0_mtp0_32.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp8/1k1k/stp/ctx1_gen3_tp8_batch1024_eplb0_mtp0_32.yaml"
-        decode:
-          num-worker: 3
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [128]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp8/1k1k/stp/ctx1_gen3_tp8_batch1024_eplb0_mtp0_128.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp8/1k1k/stp/ctx1_gen3_tp8_batch1024_eplb0_mtp0_128.yaml"
-        decode:
-          num-worker: 3
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # Non-MTP (STP) configurations - High throughput (DP attention)
-      - conc-list: [1920]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp8/1k1k/stp/ctx1_gen5_dep8_batch48_eplb0_mtp0_1920.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp8/1k1k/stp/ctx1_gen5_dep8_batch48_eplb0_mtp0_1920.yaml"
-        decode:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [4096]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp8/1k1k/stp/ctx1_gen1_dep8_batch512_eplb0_mtp0_4096.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp8/1k1k/stp/ctx1_gen1_dep8_batch512_eplb0_mtp0_4096.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [5152]
-        prefill:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b200-fp8/1k1k/stp/ctx2_gen5_dep8_batch128_eplb0_mtp0_5152.yaml
-          - "CONFIG_FILE=recipes/trtllm/b200-fp8/1k1k/stp/ctx2_gen5_dep8_batch128_eplb0_mtp0_5152.yaml"
-        decode:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-dsr1-fp4-b300-dynamo-trt:
-  image: nvcr.io/nvidia/ai-dynamo/tensorrtllm-runtime:0.8.1.post1
-  model: deepseek-r1-fp4
-  model-prefix: dsr1
-  runner: b300
-  precision: fp4
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "0.8.1.post1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - spec-decoding: "mtp"
-        conc-list: [654]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp4/1k1k/mtp/ctx1_gen1_dep8_batch64_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp4/1k1k/mtp/ctx1_gen1_dep8_batch64_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [271]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp4/1k1k/mtp/ctx1_gen2_dep8_batch16_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp4/1k1k/mtp/ctx1_gen2_dep8_batch16_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [11]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp4/1k1k/mtp/ctx1_gen5_tep8_batch1_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp4/1k1k/mtp/ctx1_gen5_tep8_batch1_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [10, 20, 25, 60, 120, 200]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp4/1k1k/mtp/ctx1_gen5_tep8_batch32_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp4/1k1k/mtp/ctx1_gen5_tep8_batch32_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [2342]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp4/1k1k/mtp/ctx2_gen1_dep8_batch256_eplb0_mtp1.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp4/1k1k/mtp/ctx2_gen1_dep8_batch256_eplb0_mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [8609]
-        prefill:
-          num-worker: 5
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp4/1k1k/mtp/ctx5_gen2_dep8_batch512_eplb0_mtp1.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp4/1k1k/mtp/ctx5_gen2_dep8_batch512_eplb0_mtp1.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [12926]
-        prefill:
-          num-worker: 5
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp4/1k1k/mtp/ctx5_gen2_dep8_batch768_eplb0_mtp1.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp4/1k1k/mtp/ctx5_gen2_dep8_batch768_eplb0_mtp1.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-      # Non-MTP configurations
-      - conc-list: [1176]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp4/1k1k/stp/ctx1_gen2_dep8_batch64_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp4/1k1k/stp/ctx1_gen2_dep8_batch64_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [6]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp4/1k1k/stp/ctx1_gen4_tep8_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp4/1k1k/stp/ctx1_gen4_tep8_batch1_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [5, 10, 15, 25]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp4/1k1k/stp/ctx1_gen5_tep4_batch4_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp4/1k1k/stp/ctx1_gen5_tep4_batch4_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [60, 110, 195, 395]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp4/1k1k/stp/ctx1_gen5_tep8_batch64_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp4/1k1k/stp/ctx1_gen5_tep8_batch64_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [4405]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp4/1k1k/stp/ctx2_gen1_dep8_batch512_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp4/1k1k/stp/ctx2_gen1_dep8_batch512_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [8192]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp4/1k1k/stp/ctx3_gen1_dep8_batch1024_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp4/1k1k/stp/ctx3_gen1_dep8_batch1024_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [4611]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp4/1k1k/stp/ctx3_gen2_dep8_batch256_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp4/1k1k/stp/ctx3_gen2_dep8_batch256_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-dsr1-fp8-b300-dynamo-trt:
-  image: nvcr.io/nvidia/ai-dynamo/tensorrtllm-runtime:0.8.1.post1
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: b300
-  precision: fp8
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "0.8.1.post1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - spec-decoding: "mtp"
-        conc-list: [10]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp8/1k1k/mtp/ctx1_gen8_tp8_batch1_eplb0_mtp3_10.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp8/1k1k/mtp/ctx1_gen8_tp8_batch1_eplb0_mtp3_10.yaml"
-        decode:
-          num-worker: 8
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [160]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp8/1k1k/mtp/ctx1_gen8_tp8_batch16_eplb0_mtp3_160.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp8/1k1k/mtp/ctx1_gen8_tp8_batch16_eplb0_mtp3_160.yaml"
-        decode:
-          num-worker: 8
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [3072]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp8/1k1k/mtp/ctx1_gen1_dp8_batch256_eplb0_mtp1_3072.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp8/1k1k/mtp/ctx1_gen1_dp8_batch256_eplb0_mtp1_3072.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [2560]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp8/1k1k/mtp/ctx1_gen2_dep8_batch128_eplb0_mtp1_2560.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp8/1k1k/mtp/ctx1_gen2_dep8_batch128_eplb0_mtp1_2560.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [720]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp8/1k1k/mtp/ctx1_gen5_dep8_batch16_eplb0_mtp2_720.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp8/1k1k/mtp/ctx1_gen5_dep8_batch16_eplb0_mtp2_720.yaml"
-        decode:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [11264]
-        prefill:
-          num-worker: 3
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp8/1k1k/mtp/ctx3_gen2_dp8_batch512_eplb0_mtp1_11264.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp8/1k1k/mtp/ctx3_gen2_dp8_batch512_eplb0_mtp1_11264.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: true
-    # 1k1k STP configs
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - conc-list: [2112]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp8/1k1k/stp/ctx1_gen1_dep8_batch256_eplb0_mtp0_2112.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp8/1k1k/stp/ctx1_gen1_dep8_batch256_eplb0_mtp0_2112.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [3072]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp8/1k1k/stp/ctx1_gen2_dp8_batch128_eplb0_mtp0_3072.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp8/1k1k/stp/ctx1_gen2_dp8_batch128_eplb0_mtp0_3072.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: true
-      - conc-list: [1280]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp8/1k1k/stp/ctx1_gen3_dp8_batch48_eplb0_mtp0_1280.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp8/1k1k/stp/ctx1_gen3_dp8_batch48_eplb0_mtp0_1280.yaml"
-        decode:
-          num-worker: 3
-          tp: 8
-          ep: 1
-          dp-attn: true
-      - conc-list: [12]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp8/1k1k/stp/ctx1_gen8_tp8_batch64_eplb0_mtp0_12.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp8/1k1k/stp/ctx1_gen8_tp8_batch64_eplb0_mtp0_12.yaml"
-        decode:
-          num-worker: 8
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [128]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp8/1k1k/stp/ctx1_gen8_tp8_batch64_eplb0_mtp0_128.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp8/1k1k/stp/ctx1_gen8_tp8_batch64_eplb0_mtp0_128.yaml"
-        decode:
-          num-worker: 8
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [384]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp8/1k1k/stp/ctx1_gen8_tp8_batch64_eplb0_mtp0_384.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp8/1k1k/stp/ctx1_gen8_tp8_batch64_eplb0_mtp0_384.yaml"
-        decode:
-          num-worker: 8
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [16384]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/b300-fp8/1k1k/stp/ctx2_gen1_dp8_batch1024_eplb0_mtp0_16384.yaml
-          - "CONFIG_FILE=recipes/trtllm/b300-fp8/1k1k/stp/ctx2_gen1_dp8_batch1024_eplb0_mtp0_16384.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-    # 8k1k MTP configs
-
-dsr1-fp4-b200-sglang:
-  image: lmsysorg/sglang:v0.5.12.post1
-  model: nvidia/DeepSeek-R1-0528-FP4-V2
-  model-prefix: dsr1
-  runner: b200
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, ep: 1, conc-start: 1, conc-end: 32 }
-      - { tp: 4, ep: 4, conc-start: 64, conc-end: 256 }
-
-dsr1-fp4-b200-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.12.post1
-  model: nvidia/DeepSeek-R1-0528-FP4-V2
-  model-prefix: dsr1
-  runner: b200
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, ep: 1, conc-start: 1, conc-end: 32, spec-decoding: mtp }
-      - { tp: 4, ep: 4, conc-start: 64, conc-end: 256, spec-decoding: mtp }
-
-# Original config key: dsv4-fp4-b200-sglang
-dsv4-fp4-b200-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260628-da802ddc
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: b200-dsv4
-  precision: fp4
-  framework: sglang
-  multinode: false
-  # Two recipes from https://docs.sglang.io/cookbook/autoregressive/DeepSeek/DeepSeek-V4
-  # are selected inside benchmarks/single_node/dsv4_fp4_b200.sh by DP_ATTENTION:
-  #   low-latency  (DP_ATTENTION=false): TP-only, flashinfer_mxfp4
-  #   DP-attention  (DP_ATTENTION=true):  DP-attn + DeepEP + mega_moe opts
-  # The DP-attention recipe covers both "balanced" (conc 64-128) and
-  # "max-throughput" (conc 256+) CONC ranges with identical flags;
-  # only --max-running-requests scales with CONC.
-  # ep is implicit in sglang: --moe-a2a-backend deepep forces ep_size=tp_size,
-  # while low-latency leaves ep_size at the default of 1.
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # low-latency (DP_ATTENTION=false)
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 32 }
-      # DP-attention (DP_ATTENTION=true) — balanced CONC range
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 64, conc-end: 128 }
-      # DP-attention (DP_ATTENTION=true) — max-throughput CONC range
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 1024 }
-
-# Original config key: dsv4-fp4-b200-vllm
-dsv4-fp4-b200-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai:v0.25.0
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: cluster:b200-dgxc
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 64 }
-      - { tp: 8, ep: 8, conc-start: 128, conc-end: 256 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 1024 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 4096, conc-end: 4096 }
-
-# Original config key: dsv4-fp4-b200-trt
-dsv4-fp4-b200-trt-deprecated-1k1k:
-  image: ghcr.io#semianalysisai/trtllm-deepseek-v4:feat-deepseek_v4-c185066
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: b200-dsv4
-  precision: fp4
-  framework: trt
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 32 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 32, conc-end: 512 }
-
-# Original config key: dsv4-fp4-b200-trt-mtp
-dsv4-fp4-b200-trt-mtp-deprecated-1k1k:
-  image: ghcr.io#semianalysisai/trtllm-deepseek-v4:feat-deepseek_v4-c185066
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: b200-dsv4
-  precision: fp4
-  framework: trt
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 32, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 32, conc-end: 256, spec-decoding: mtp }
-
-# Original config key: dsv4-fp4-b200-vllm-mtp
-dsv4-fp4-b200-vllm-mtp-deprecated-1k1k:
-  image: vllm/vllm-openai:v0.25.0
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: cluster:b200-dgxc
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 8, ep: 8, conc-start: 128, conc-end: 256, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 4096, spec-decoding: mtp }
-
-dsr1-fp4-b300-sglang:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: nvidia/DeepSeek-R1-0528-FP4-V2
-  model-prefix: dsr1
-  runner: b300
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, ep: 4, conc-start: 1, conc-end: 128 }
-      - { tp: 8, ep: 8, conc-start: 1, conc-end: 128 }
-
-dsr1-fp4-b200-trt:
-  image: nvcr.io#nvidia/tensorrt-llm/release:1.3.0rc14
-  model: nvidia/DeepSeek-R1-0528-FP4-V2
-  model-prefix: dsr1
-  runner: b200
-  precision: fp4
-  framework: trt
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # low concurrency cases use TP only
-      # concurrency 64 uses TP & EP
-      # high concurrency cases use TP & EP & DP-ATTN
-      - { tp: 4, conc-start: 4, conc-end: 16 }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 256, conc-end: 256 }
-      - { tp: 8, conc-start: 4, conc-end: 4 }
-      - { tp: 8, ep: 8, conc-start: 64, conc-end: 64 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 128, conc-end: 256 }
-
-dsr1-fp4-b200-trt-mtp:
-  image: nvcr.io#nvidia/tensorrt-llm/release:1.3.0rc14
-  model: nvidia/DeepSeek-R1-0528-FP4-V2
-  model-prefix: dsr1
-  runner: b200
-  precision: fp4
-  framework: trt
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # TP=4 configurations
-      - { tp: 4, conc-start: 4, conc-end: 8, spec-decoding: mtp }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 256, conc-end: 256, spec-decoding: mtp }
-      # TP=8 configurations
-      - { tp: 8, conc-start: 4, conc-end: 4, spec-decoding: mtp }
-      - { tp: 8, conc-start: 128, conc-end: 128, spec-decoding: mtp }
-      - { tp: 8, ep: 8, conc-start: 32, conc-end: 128, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 32, conc-end: 64, spec-decoding: mtp }
-
-dsr1-fp8-b200-sglang:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: b200
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 64 }
-
-dsr1-fp8-b300-sglang:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: b300
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 64 }
-
-# Original config key: dsv4-fp4-b300-sglang
-dsv4-fp4-b300-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260624-b2c8f7a2
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: b300
-  precision: fp4
-  framework: sglang
-  multinode: false
-  # Recipes are selected inside benchmarks/single_node/dsv4_fp4_b300_sglang.sh
-  # by CONC:
-  #   CONC 1|32:         TP-only, flashinfer_mxfp4
-  #   CONC 512:          DP-attn, flashinfer_mxfp4
-  #   CONC 2048-8192:    DP-attn, megamoe
-  # ep is implicit in sglang: --moe-a2a-backend megamoe forces ep_size=tp_size,
-  # while low-latency leaves ep_size at the default of 1.
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 1 }
-      - { tp: 4, ep: 1, conc-start: 32, conc-end: 32 }
-      - { tp: 4, ep: 1, dp-attn: true, conc-start: 512, conc-end: 512 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 8192, conc-end: 8192 }
-
-# Original config key: dsv4-fp4-b300-sglang-mtp
-dsv4-fp4-b300-sglang-mtp-deprecated-1k1k:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260610-f332e526
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: b300
-  precision: fp4
-  framework: sglang
-  multinode: false
-  # Three CONC bands:
-  #   A: TP=8 ep=1            -- conc 1-8    EAGLE (3,1,4) TP-only fallback
-  #   B: TP=4 ep=1            -- conc 4-32   EAGLE (3,1,4) TP-only mid batch
-  #   C: TP=4 ep=1 dp-attn    -- conc 16-256 EAGLE (1,1,2) DP-attn flashinfer
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 8, spec-decoding: mtp }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 32, spec-decoding: mtp }
-
-# Original config key: qwen3.5-bf16-b200-sglang
-qwen3.5-bf16-b200-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.14-cu130
-  model: Qwen/Qwen3.5-397B-A17B
-  model-prefix: qwen3.5
-  runner: b200
-  precision: bf16
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64 }
-
-# Original config key: qwen3.5-bf16-b200-sglang-mtp
-qwen3.5-bf16-b200-sglang-mtp-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.14-cu130
-  model: Qwen/Qwen3.5-397B-A17B
-  model-prefix: qwen3.5
-  runner: b200
-  precision: bf16
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-
-qwen3.5-fp8-b200-sglang:
-  image: lmsysorg/sglang:v0.5.14-cu130
-  model: Qwen/Qwen3.5-397B-A17B-FP8
-  model-prefix: qwen3.5
-  runner: b200
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 4 }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 }
-
-qwen3.5-fp4-b200-sglang:
-  image: lmsysorg/sglang:v0.5.14-cu130
-  model: nvidia/Qwen3.5-397B-A17B-NVFP4
-  model-prefix: qwen3.5
-  runner: b200
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 4 }
-      - { tp: 2, ep: 1, conc-start: 4, conc-end: 128 }
-
-qwen3.5-fp4-b200-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.14-cu130
-  model: nvidia/Qwen3.5-397B-A17B-NVFP4
-  model-prefix: qwen3.5
-  runner: b200
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 4, spec-decoding: mtp }
-      - { tp: 2, ep: 1, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-
-# Original config key: glm5-fp8-b200-sglang
-glm5-fp8-b200-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260605-7dc73766
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: b200
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 256 }
-
-# Original config key: glm5-fp8-b200-sglang-mtp
-glm5-fp8-b200-sglang-mtp-deprecated-1k1k:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260605-7dc73766
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: b200
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 256, spec-decoding: mtp }
-
-# Original config key: glm5-fp8-b300-sglang
-glm5-fp8-b300-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: b300
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 256 }
-
-# Original config key: glm5-fp8-b300-sglang-mtp
-glm5-fp8-b300-sglang-mtp-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: b300
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 256, spec-decoding: mtp }
-
-# Original config key: glm5-fp4-b200-sglang
-glm5-fp4-b200-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260605-7dc73766
-  model: nvidia/GLM-5-NVFP4
-  model-prefix: glm5
-  runner: b200
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 4 }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 }
-
-# Original config key: glm5-fp4-b200-sglang-mtp
-glm5-fp4-b200-sglang-mtp-deprecated-1k1k:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260605-7dc73766
-  model: nvidia/GLM-5-NVFP4
-  model-prefix: glm5
-  runner: b200
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 4, spec-decoding: mtp }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 256, spec-decoding: mtp }
-
-# Original config key: glm5-fp4-b300-sglang
-glm5-fp4-b300-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.11-cu130
-  model: nvidia/GLM-5-NVFP4
-  model-prefix: glm5
-  runner: b300
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 4 }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 }
-
-# Original config key: glm5-fp4-b300-sglang-mtp
-glm5-fp4-b300-sglang-mtp-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.11-cu130
-  model: nvidia/GLM-5-NVFP4
-  model-prefix: glm5
-  runner: b300
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 4, spec-decoding: mtp }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 256, spec-decoding: mtp }
-
-# Original config key: glm5-fp4-gb300-dynamo-trt
-glm5-fp4-gb300-dynamo-trt-deprecated-1k1k:
-  image: nvcr.io/nvidia/ai-dynamo/tensorrtllm-runtime:1.3.0-dev.1-cuda13
-  model: nvidia/GLM-5-NVFP4
-  model-prefix: glm5
-  runner: gb300-nv
-  precision: fp4
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "v1.3.0-dev.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # STP configurations
-      - conc-list: [ 4 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx1dep2_gen4tep8_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx1dep2_gen4tep8_batch1_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 5 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx1dep2_gen5tep4_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx1dep2_gen5tep4_batch1_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 24 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx1dep2_gen4tep8_batch4_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx1dep2_gen4tep8_batch4_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 92 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx1dep2_gen4tep8_batch16_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx1dep2_gen4tep8_batch16_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 105 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx1dep2_gen5tep4_batch16_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx1dep2_gen5tep4_batch16_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 336 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx1dep2_gen4tep8_batch64_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx1dep2_gen4tep8_batch64_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 666 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx1dep2_gen1dep32_batch16_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx1dep2_gen1dep32_batch16_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 2253 ]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx2dep2_gen1dep16_batch128_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx2dep2_gen1dep16_batch128_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-      - conc-list: [ 2253 ]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx3dep2_gen1dep32_batch64_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx3dep2_gen1dep32_batch64_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 4301 ]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx3dep2_gen1dep16_batch256_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx3dep2_gen1dep16_batch256_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [ 8192 ]
-        prefill:
-          num-worker: 4
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx4dep2_gen1dep16_batch512_eplb256_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/STP/ctx4dep2_gen1dep16_batch512_eplb256_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-# Original config key: glm5-fp4-gb200-dynamo-trt
-glm5-fp4-gb200-dynamo-trt-deprecated-1k1k:
-  image: nvcr.io/nvidia/ai-dynamo/tensorrtllm-runtime:1.3.0-dev.1-cuda13
-  model: nvidia/GLM-5-NVFP4
-  model-prefix: glm5
-  runner: gb200
-  precision: fp4
-  framework: dynamo-trt
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # STP configurations
-      - conc-list: [ 4 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p4d-dep4-tep8-b1-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 5 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen5tep4_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p5d-dep4-tep4-b1-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 20 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch4_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p4d-dep4-tep8-b4-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 25 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen5tep4_batch4_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p5d-dep4-tep4-b4-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 84 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch16_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p4d-dep4-tep8-b16-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 168 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch32_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p4d-dep4-tep8-b32-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 284 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch64_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p4d-dep4-tep8-b64-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 666 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch16_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p1d-dep4-dep32-b16-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 1229 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch32_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p1d-dep4-dep32-b32-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 2151 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep32_batch64_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-2p1d-dep4-dep32-b64-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 2151 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep16_batch128_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-2p1d-dep4-dep16-b128-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [ 4301 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep16_batch256_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-2p1d-dep4-dep16-b256-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [ 4301 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep8_batch512_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p1d-dep4-dep8-b512-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-# Original config key: glm5-fp4-gb200-dynamo-trt-mtp
-glm5-fp4-gb200-dynamo-trt-mtp-deprecated-1k1k:
-  image: nvcr.io/nvidia/ai-dynamo/tensorrtllm-runtime:1.3.0-dev.1-cuda13
-  model: nvidia/GLM-5-NVFP4
-  model-prefix: glm5
-  runner: gb200
-  precision: fp4
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "v1.3.0-dev.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # MTP configurations
-      - spec-decoding: "mtp"
-        conc-list: [ 8 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen4tep8_batch1_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p4d-dep4-tep8-b1-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [ 10 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen5tep4_batch1_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p5d-dep4-tep4-b1-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [ 24 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen4tep8_batch4_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p4d-dep4-tep8-b4-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [ 40 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen4tep8_batch8_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p4d-dep4-tep8-b8-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [ 92 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen4tep8_batch16_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p4d-dep4-tep8-b16-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [ 180 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen1dep32_batch4_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p1d-dep4-dep32-b4-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 308 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen1dep32_batch8_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p1d-dep4-dep32-b8-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 615 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen1dep16_batch32_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p1d-dep4-dep16-b32-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 666 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen1dep32_batch16_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p1d-dep4-dep32-b16-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 1229 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen1dep16_batch64_eplb0_mtp1.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-1p1d-dep4-dep16-b64-eplb0-mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 2253 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/MTP/ctx2dep4_gen1dep16_batch128_eplb0_mtp1.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-2p1d-dep4-dep16-b128-eplb0-mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 4301 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/MTP/ctx2dep4_gen1dep16_batch256_eplb256_mtp1.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-2p1d-dep4-dep16-b256-eplb256-mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 4301 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL1K_OSL1K/MTP/ctx2dep4_gen1dep8_batch512_eplb0_mtp1.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/1k1k/disagg-2p1d-dep4-dep8-b512-eplb0-mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-qwen3.5-fp8-b200-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.14-cu130
-  model: Qwen/Qwen3.5-397B-A17B-FP8
-  model-prefix: qwen3.5
-  runner: b200
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 4, spec-decoding: mtp }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 256, spec-decoding: mtp }
-
-qwen3.5-fp8-b300-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: Qwen/Qwen3.5-397B-A17B-FP8
-  model-prefix: qwen3.5
-  runner: b300
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 256, spec-decoding: mtp }
-
-qwen3.5-fp8-b300-sglang:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: Qwen/Qwen3.5-397B-A17B-FP8
-  model-prefix: qwen3.5
-  runner: b300
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 }
-
-qwen3.5-fp4-b300-sglang:
-  image: lmsysorg/sglang:v0.5.14-cu130
-  model: nvidia/Qwen3.5-397B-A17B-NVFP4
-  model-prefix: qwen3.5
-  runner: b300
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 128 }
-      - { tp: 2, ep: 2, conc-start: 4, conc-end: 128 }
-
-qwen3.5-fp4-b300-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.14-cu130
-  model: nvidia/Qwen3.5-397B-A17B-NVFP4
-  model-prefix: qwen3.5
-  runner: b300
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 128, spec-decoding: mtp }
-      - { tp: 2, ep: 2, conc-start: 4, conc-end: 128, spec-decoding: mtp }
-
-# Original config key: qwen3.5-bf16-b300-sglang
-qwen3.5-bf16-b300-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: Qwen/Qwen3.5-397B-A17B
-  model-prefix: qwen3.5
-  runner: b300
-  precision: bf16
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64 }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 64 }
-
-# Original config key: qwen3.5-bf16-b300-sglang-mtp
-qwen3.5-bf16-b300-sglang-mtp-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: Qwen/Qwen3.5-397B-A17B
-  model-prefix: qwen3.5
-  runner: b300
-  precision: bf16
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-
-# Original config key: kimik2.5-int4-b200-vllm
-kimik2.5-int4-b200-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai:v0.24.0
-  model: moonshotai/Kimi-K2.5
-  model-prefix: kimik2.5
-  runner: b200
-  precision: int4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-# Original config key: kimik2.5-int4-b300-vllm
-kimik2.5-int4-b300-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai:v0.24.0
-  model: moonshotai/Kimi-K2.5
-  model-prefix: kimik2.5
-  runner: b300
-  precision: int4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 64 }
-
-# Original config key: kimik2.5-int4-h200-vllm
-kimik2.5-int4-h200-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai:v0.22.0
-  model: moonshotai/Kimi-K2.5
-  model-prefix: kimik2.5
-  runner: h200
-  precision: int4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-# Original config key: kimik2.5-fp4-b200-vllm
-kimik2.5-fp4-b200-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai:v0.22.0
-  model: nvidia/Kimi-K2.5-NVFP4
-  model-prefix: kimik2.5
-  runner: b200
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 4 }
-      - { tp: 4, ep: 1, conc-start: 1, conc-end: 128 }
-
-# Original config key: kimik2.5-fp4-b300-vllm
-kimik2.5-fp4-b300-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai:v0.22.0
-  model: nvidia/Kimi-K2.5-NVFP4
-  model-prefix: kimik2.5
-  runner: b300
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 4 }
-      - { tp: 4, ep: 1, conc-start: 1, conc-end: 128 }
-
-dsr1-fp8-b200-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: b200
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 512, spec-decoding: mtp }
-
-dsr1-fp8-b300-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: b300
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 512, spec-decoding: mtp }
-
-dsr1-fp8-b200-trt:
-  image: nvcr.io#nvidia/tensorrt-llm/release:1.3.0rc14
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: b200
-  precision: fp8
-  framework: trt
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 64, conc-end: 128 }
-      - { tp: 4, ep: 1, conc-start: 8, conc-end: 16 }
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 8 }
-
-dsr1-fp8-b200-trt-mtp:
-  image: nvcr.io#nvidia/tensorrt-llm/release:1.3.0rc14
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: b200
-  precision: fp8
-  framework: trt
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # mostly TP8
-      # If CONC == 256, then TP8, EP8, DP_ATTN=true
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 128, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 256, spec-decoding: mtp }
-
-dsr1-fp8-h200-sglang:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: h200
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-dsr1-fp8-h200-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: h200
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-
-# Original config key: dsv4-fp8-h200-vllm
-dsv4-fp8-h200-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai:v0.25.1
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: h200
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, dp-attn: false, conc-start: 1, conc-end: 256 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 1, conc-end: 256 }
-
-# Original config key: dsv4-fp8-h200-vllm-mtp
-dsv4-fp8-h200-vllm-mtp-deprecated-1k1k:
-  image: vllm/vllm-openai:v0.25.1
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: h200
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, dp-attn: false, conc-start: 1, conc-end: 256, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 1, conc-end: 256, spec-decoding: mtp }
-
-# Original config key: dsv4-fp8-h200-sglang
-dsv4-fp8-h200-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang:deepseek-v4-hopper@sha256:1bf5d508ab110cc0fe1659a5f21d1be02a7f0d7ca8f58cea7e7f4e11f6ae208f
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: h200-dgxc
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 1 }
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64 }
-
-# Original config key: dsv4-fp8-h200-sglang-mtp
-dsv4-fp8-h200-sglang-mtp-deprecated-1k1k:
-  image: lmsysorg/sglang:deepseek-v4-hopper@sha256:1bf5d508ab110cc0fe1659a5f21d1be02a7f0d7ca8f58cea7e7f4e11f6ae208f
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: h200-dgxc
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 1, spec-decoding: mtp }
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-
-# Original config key: dsv4-fp4-b300-vllm
-dsv4-fp4-b300-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai:v0.25.0
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: cluster:b300-nv
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 128 }
-      - { tp: 8, conc-start: 1, conc-end: 4 }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 256, conc-end: 512 }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 2048, conc-end: 2048 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 4096, conc-end: 8192 }
-
-# Original config key: dsv4-fp4-b300-trt
-dsv4-fp4-b300-trt-deprecated-1k1k:
-  image: ghcr.io#semianalysisai/trtllm-deepseek-v4:feat-deepseek_v4-c185066
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: b300
-  precision: fp4
-  framework: trt
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 64 }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 64, conc-end: 256 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 512, conc-end: 1024 }
-
-# Original config key: dsv4-fp4-b300-trt-mtp
-dsv4-fp4-b300-trt-mtp-deprecated-1k1k:
-  image: ghcr.io#semianalysisai/trtllm-deepseek-v4:feat-deepseek_v4-c185066
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: b300
-  precision: fp4
-  framework: trt
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 64, conc-end: 256, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 512, conc-end: 1024, spec-decoding: mtp }
-
-# Original config key: dsv4-fp4-b300-vllm-mtp
-dsv4-fp4-b300-vllm-mtp-deprecated-1k1k:
-  image: vllm/vllm-openai:v0.25.0
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: cluster:b300-nv
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 256, spec-decoding: mtp }
-      - { tp: 8, conc-start: 1, conc-end: 8, spec-decoding: mtp }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 256, conc-end: 1024, spec-decoding: mtp }
-
-qwen3.5-fp8-h200-sglang:
-  image: lmsysorg/sglang:v0.5.14-cu130
-  model: Qwen/Qwen3.5-397B-A17B-FP8
-  model-prefix: qwen3.5
-  runner: h200
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 8, conc-start: 4, conc-end: 64 }
-
-qwen3.5-fp8-h200-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.14-cu130
-  model: Qwen/Qwen3.5-397B-A17B-FP8
-  model-prefix: qwen3.5
-  runner: h200
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 8, conc-start: 4, conc-end: 128, spec-decoding: mtp }
-
-# Original config key: glm5-fp8-h200-sglang
-glm5-fp8-h200-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: h200-dgxc
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-# Original config key: glm5-fp8-h200-sglang-mtp
-glm5-fp8-h200-sglang-mtp-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: h200
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-
-dsr1-fp8-h200-trt:
-  image: nvcr.io#nvidia/tensorrt-llm/release:1.3.0rc14
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: h200
-  precision: fp8
-  framework: trt
-  multinode: false
-  # For all sequence lengths, EP=TP
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      # If CONC > 64, then DP_ATTN=true
-      search-space:
-      - { tp: 8, ep: 8, conc-start: 4, conc-end: 64 }
-
-dsr1-fp8-h200-trt-mtp:
-  image: nvcr.io#nvidia/tensorrt-llm/release:1.3.0rc14
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: h200
-  precision: fp8
-  framework: trt
-  multinode: false
-  # For all sequence lengths, EP=TP, MOE_BACKEND=CUTLASS, MTP=3 (or MTP=1 when DP_ATTN=true)
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # If CONC >= 128, then DP_ATTN=true, MTP=1
-      - { tp: 8, ep: 8, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 128, conc-end: 256, spec-decoding: mtp }
-
-dsr1-fp8-h200-dynamo-trt:
-  image: nvcr.io/nvidia/ai-dynamo/tensorrtllm-runtime:0.8.1.post1
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: cluster:h200-dgxc
-  precision: fp8
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "0.8.1.post1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # MTP configurations
-      - spec-decoding: "mtp"
-        conc-list: [1]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h200/1k1k/mtp/c1_ctx1_gen11_tep8_batch1_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/h200/1k1k/mtp/c1_ctx1_gen11_tep8_batch1_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 11
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [4]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h200/1k1k/mtp/c4_ctx1_gen11_tep8_batch128_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/h200/1k1k/mtp/c4_ctx1_gen11_tep8_batch128_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 11
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [8]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h200/1k1k/mtp/c8_ctx1_gen11_tep8_batch128_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/h200/1k1k/mtp/c8_ctx1_gen11_tep8_batch128_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 11
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [16]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h200/1k1k/mtp/c16_ctx1_gen9_tep8_batch128_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/h200/1k1k/mtp/c16_ctx1_gen9_tep8_batch128_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 9
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [32]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h200/1k1k/mtp/c32_ctx1_gen11_tep8_batch128_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/h200/1k1k/mtp/c32_ctx1_gen11_tep8_batch128_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 11
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [64]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h200/1k1k/mtp/c64_ctx1_gen8_dep8_batch128_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/h200/1k1k/mtp/c64_ctx1_gen8_dep8_batch128_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 8
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [128]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h200/1k1k/mtp/c128_ctx1_gen7_dep8_batch128_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/h200/1k1k/mtp/c128_ctx1_gen7_dep8_batch128_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 7
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [256]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h200/1k1k/mtp/c256_ctx1_gen4_dep8_batch128_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/h200/1k1k/mtp/c256_ctx1_gen4_dep8_batch128_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [512]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h200/1k1k/mtp/c512_ctx1_gen2_dep8_batch256_eplb0_mtp1.yaml
-          - "CONFIG_FILE=recipes/trtllm/h200/1k1k/mtp/c512_ctx1_gen2_dep8_batch256_eplb0_mtp1.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # Non-MTP configurations (STP)
-      - conc-list: [1]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h200/1k1k/stp/c1_ctx1_gen9_tep8_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/h200/1k1k/stp/c1_ctx1_gen9_tep8_batch1_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 9
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [4]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h200/1k1k/stp/c4_ctx1_gen9_tep8_batch256_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/h200/1k1k/stp/c4_ctx1_gen9_tep8_batch256_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 9
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [8]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h200/1k1k/stp/c8_ctx1_gen9_tep8_batch256_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/h200/1k1k/stp/c8_ctx1_gen9_tep8_batch256_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 9
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [16]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h200/1k1k/stp/c16_ctx1_gen9_tep8_batch256_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/h200/1k1k/stp/c16_ctx1_gen9_tep8_batch256_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 9
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [32]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h200/1k1k/stp/c32_ctx1_gen9_tep8_batch256_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/h200/1k1k/stp/c32_ctx1_gen9_tep8_batch256_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 9
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [64]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h200/1k1k/stp/c64_ctx1_gen9_tep8_batch256_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/h200/1k1k/stp/c64_ctx1_gen9_tep8_batch256_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 9
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [128]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h200/1k1k/stp/c128_ctx1_gen9_dep8_batch512_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/h200/1k1k/stp/c128_ctx1_gen9_dep8_batch512_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 9
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [256]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h200/1k1k/stp/c256_ctx1_gen6_dep8_batch512_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/h200/1k1k/stp/c256_ctx1_gen6_dep8_batch512_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 6
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [512]
-        prefill:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h200/1k1k/stp/c512_ctx2_gen7_dep8_batch512_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/h200/1k1k/stp/c512_ctx2_gen7_dep8_batch512_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 7
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-dsr1-fp8-h100-dynamo-trt:
-  image: nvcr.io/nvidia/ai-dynamo/tensorrtllm-runtime:0.8.1.post3
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: cluster:h100-dgxc
-  precision: fp8
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "0.8.1.post3" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # MTP configurations
-      - spec-decoding: "mtp"
-        conc-list: [6]
-        prefill:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h100-fp8/1k1k/mtp/ctx1_gen3_tep16_batch1_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/h100-fp8/1k1k/mtp/ctx1_gen3_tep16_batch1_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 3
-          tp: 16
-          ep: 16
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [9]
-        prefill:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h100-fp8/1k1k/mtp/ctx1_gen3_tep16_batch2_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/h100-fp8/1k1k/mtp/ctx1_gen3_tep16_batch2_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 3
-          tp: 16
-          ep: 16
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [30]
-        prefill:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h100-fp8/1k1k/mtp/ctx1_gen3_tep16_batch8_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/h100-fp8/1k1k/mtp/ctx1_gen3_tep16_batch8_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 3
-          tp: 16
-          ep: 16
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [60]
-        prefill:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h100-fp8/1k1k/mtp/ctx1_gen3_tep16_batch16_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/h100-fp8/1k1k/mtp/ctx1_gen3_tep16_batch16_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 3
-          tp: 16
-          ep: 16
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [117]
-        prefill:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h100-fp8/1k1k/mtp/ctx1_gen3_tep16_batch32_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/h100-fp8/1k1k/mtp/ctx1_gen3_tep16_batch32_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 3
-          tp: 16
-          ep: 16
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [231]
-        prefill:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h100-fp8/1k1k/mtp/ctx1_gen3_dep16_batch4_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/h100-fp8/1k1k/mtp/ctx1_gen3_dep16_batch4_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 3
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [462]
-        prefill:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h100-fp8/1k1k/mtp/ctx1_gen3_tep16_batch128_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/h100-fp8/1k1k/mtp/ctx1_gen3_tep16_batch128_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 3
-          tp: 16
-          ep: 16
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [615]
-        prefill:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h100-fp8/1k1k/mtp/ctx1_gen1_dep16_batch32_eplb0_mtp2.yaml
-          - "CONFIG_FILE=recipes/trtllm/h100-fp8/1k1k/mtp/ctx1_gen1_dep16_batch32_eplb0_mtp2.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [1229]
-        prefill:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h100-fp8/1k1k/mtp/ctx1_gen1_dep16_batch64_eplb0_mtp1.yaml
-          - "CONFIG_FILE=recipes/trtllm/h100-fp8/1k1k/mtp/ctx1_gen1_dep16_batch64_eplb0_mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # Non-MTP configurations (STP)
-      - conc-list: [6]
-        prefill:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h100-fp8/1k1k/stp/ctx1_gen3_tep16_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/h100-fp8/1k1k/stp/ctx1_gen3_tep16_batch1_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 3
-          tp: 16
-          ep: 16
-          dp-attn: false
-      - conc-list: [9]
-        prefill:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h100-fp8/1k1k/stp/ctx1_gen3_tep16_batch2_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/h100-fp8/1k1k/stp/ctx1_gen3_tep16_batch2_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 3
-          tp: 16
-          ep: 16
-          dp-attn: false
-      - conc-list: [30]
-        prefill:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h100-fp8/1k1k/stp/ctx1_gen3_tep16_batch8_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/h100-fp8/1k1k/stp/ctx1_gen3_tep16_batch8_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 3
-          tp: 16
-          ep: 16
-          dp-attn: false
-      - conc-list: [60]
-        prefill:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h100-fp8/1k1k/stp/ctx1_gen3_tep16_batch16_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/h100-fp8/1k1k/stp/ctx1_gen3_tep16_batch16_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 3
-          tp: 16
-          ep: 16
-          dp-attn: false
-      - conc-list: [231]
-        prefill:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h100-fp8/1k1k/stp/ctx1_gen3_dep16_batch4_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/h100-fp8/1k1k/stp/ctx1_gen3_dep16_batch4_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 3
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [462]
-        prefill:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h100-fp8/1k1k/stp/ctx1_gen3_dep16_batch8_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/h100-fp8/1k1k/stp/ctx1_gen3_dep16_batch8_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 3
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [924]
-        prefill:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h100-fp8/1k1k/stp/ctx1_gen3_dep16_batch16_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/h100-fp8/1k1k/stp/ctx1_gen3_dep16_batch16_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 3
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [1845]
-        prefill:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h100-fp8/1k1k/stp/ctx1_gen3_dep16_batch32_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/h100-fp8/1k1k/stp/ctx1_gen3_dep16_batch32_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 3
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [4916]
-        prefill:
-          num-worker: 2
-          tp: 16
-          ep: 16
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/h100-fp8/1k1k/stp/ctx2_gen1_dep16_batch256_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/h100-fp8/1k1k/stp/ctx2_gen1_dep16_batch256_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-# Original config key: minimaxm3-fp8-h100-vllm
-minimaxm3-fp8-h100-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai:minimax-m3
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: h100
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # DEP (dp-attn) omitted on H100: each DP rank replicates the ~20 GB
-      # BF16-dequantized attention/dense/embedding weights next to its
-      # ~52 GB expert shard, and KV-cache init fails at high conc (sweep
-      # 27441767143, conc 256/512: "No available memory for the cache
-      # blocks"). TEP8 covers the high-concurrency regime instead.
-      - { tp: 8, conc-start: 1, conc-end: 128 }
-      - { tp: 8, ep: 8, conc-start: 1, conc-end: 256 }
-
-dsr1-fp8-h100-dynamo-sglang:
-  image: lmsysorg/sglang:v0.5.8-cu130
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: cluster:h100-dgxc
-  precision: fp8
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "0.8.0" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # # STP: Max throughput TEP (1 prefill, 2 decode)
-      # - conc-list: [1, 2, 4, 8, 16, 32, 64, 128]
-      #   prefill:
-      #     num-worker: 1
-      #     tp: 16
-      #     ep: 1
-      #     dp-attn: false
-      #     additional-settings:
-      #     - "CONFIG_FILE=recipes/h100/1k1k/stp/h100-fp8-1p2d-max-tp.yaml"
-      #   decode:
-      #     num-worker: 2
-      #     tp: 16
-      #     ep: 1
-      #     dp-attn: false
-      # # STP: Max throughput DEP (1 prefill, 1 decode, dp-attention)
-      # - conc-list: [1, 2, 4, 8, 16, 32, 64]
-      #   prefill:
-      #     num-worker: 1
-      #     tp: 16
-      #     ep: 1
-      #     dp-attn: false
-      #     additional-settings:
-      #     - "CONFIG_FILE=recipes/h100/1k1k/stp/h100-fp8-1p1d-max-dep.yaml"
-      #   decode:
-      #     num-worker: 1
-      #     tp: 16
-      #     ep: 16
-      #     dp-attn: true
-      # MTP: Max throughput TEP (1 prefill, 2 decode)
-      - spec-decoding: "mtp"
-        conc-list: [1, 2, 4, 8, 16, 32, 64, 128]
-        prefill:
-          num-worker: 1
-          tp: 16
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/h100/1k1k/mtp/h100-fp8-1p2d-max-tp-mtp.yaml"
-        decode:
-          num-worker: 2
-          tp: 16
-          ep: 1
-          dp-attn: false
-      # MTP: Max throughput DEP (1 prefill, 1 decode, dp-attention)
-      - spec-decoding: "mtp"
-        conc-list: [1, 2, 4, 8, 16, 32, 64]
-        prefill:
-          num-worker: 1
-          tp: 16
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/h100/1k1k/mtp/h100-fp8-1p1d-max-dep-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-# Original config key: minimaxm3-fp8-h200-vllm
-minimaxm3-fp8-h200-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai:minimax-m3
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: h200
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 64 }
-      - { tp: 4, ep: 4, conc-start: 1, conc-end: 256 }
-      - { tp: 8, conc-start: 1, conc-end: 128 }
-      - { tp: 8, ep: 8, conc-start: 1, conc-end: 256 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 1024 }
-
-dsr1-fp4-gb200-dynamo-trt:
-  image: nvcr.io/nvidia/ai-dynamo/tensorrtllm-runtime:0.8.1.post2
-  model: nvidia/DeepSeek-R1-0528-NVFP4-v2
-  model-prefix: dsr1
-  runner: gb200
-  precision: fp4
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "0.8.1.post2" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # MTP configurations (spec_decoding="mtp")
-      - spec-decoding: "mtp"
-        conc-list: [ 180 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp4/1k1k/mtp/ctx1_gen1_dep32_batch4_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp4/1k1k/mtp/ctx1_gen1_dep32_batch4_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 4, 8, 12, 24, 48 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp4/1k1k/mtp/ctx1_gen4_tep8_batch8_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp4/1k1k/mtp/ctx1_gen4_tep8_batch8_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [ 4301 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp4/1k1k/mtp/ctx2_gen1_dep16_batch256_eplb256_mtp1.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp4/1k1k/mtp/ctx2_gen1_dep16_batch256_eplb256_mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 2253 ]
-        prefill:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp4/1k1k/mtp/ctx3_gen1_dep32_batch64_eplb288_mtp1.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp4/1k1k/mtp/ctx3_gen1_dep32_batch64_eplb288_mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 16130 ]
-        prefill:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp4/1k1k/mtp/ctx3_gen5_dep4_batch768_eplb0_mtp1.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp4/1k1k/mtp/ctx3_gen5_dep4_batch768_eplb0_mtp1.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: true
-
-      # Non-MTP configurations (default spec_decoding="none")
-      - conc-list: [ 4301 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp4/1k1k/stp/ctx1_gen1_dep8_batch512_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp4/1k1k/stp/ctx1_gen1_dep8_batch512_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [ 666 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp4/1k1k/stp/ctx1_gen1_dep32_batch16_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp4/1k1k/stp/ctx1_gen1_dep32_batch16_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 6144 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp4/1k1k/stp/ctx1_gen2_dep4_batch768_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp4/1k1k/stp/ctx1_gen2_dep4_batch768_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-      - conc-list: [ 12, 24, 48, 96, 192 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp4/1k1k/stp/ctx1_gen4_tep8_batch32_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp4/1k1k/stp/ctx1_gen4_tep8_batch32_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 5 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp4/1k1k/stp/ctx1_gen4_tep8_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp4/1k1k/stp/ctx1_gen4_tep8_batch1_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 4301 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp4/1k1k/stp/ctx2_gen1_dep16_batch256_eplb256_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp4/1k1k/stp/ctx2_gen1_dep16_batch256_eplb256_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [ 2253 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp4/1k1k/stp/ctx2_gen1_dep32_batch64_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp4/1k1k/stp/ctx2_gen1_dep32_batch64_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-
-dsr1-fp8-gb200-dynamo-trt:
-  image: nvcr.io#nvidia/ai-dynamo/tensorrtllm-runtime:0.8.1.post2
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: gb200
-  precision: fp8
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "0.8.1.post2" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - spec-decoding: "mtp"
-        conc-list: [4301]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp8/1k1k/mtp/ctx1_gen1_dep8_batch512_eplb0_mtp1_4301.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp8/1k1k/mtp/ctx1_gen1_dep8_batch512_eplb0_mtp1_4301.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [2151]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp8/1k1k/mtp/ctx1_gen1_dep8_batch256_eplb0_mtp1_2151.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp8/1k1k/mtp/ctx1_gen1_dep8_batch256_eplb0_mtp1_2151.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [1229]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp8/1k1k/mtp/ctx1_gen1_dep16_batch64_eplb0_mtp1_1229.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp8/1k1k/mtp/ctx1_gen1_dep16_batch64_eplb0_mtp1_1229.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [615]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp8/1k1k/mtp/ctx1_gen1_dep32_batch16_eplb0_mtp3_615.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp8/1k1k/mtp/ctx1_gen1_dep32_batch16_eplb0_mtp3_615.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [36]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp8/1k1k/mtp/ctx1_gen3_tep8_batch8_eplb0_mtp3_36.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp8/1k1k/mtp/ctx1_gen3_tep8_batch8_eplb0_mtp3_36.yaml"
-        decode:
-          num-worker: 3
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [18]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp8/1k1k/mtp/ctx1_gen3_tep8_batch4_eplb0_mtp3_18.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp8/1k1k/mtp/ctx1_gen3_tep8_batch4_eplb0_mtp3_18.yaml"
-        decode:
-          num-worker: 3
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [9]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp8/1k1k/mtp/ctx1_gen3_tep8_batch2_eplb0_mtp3_9.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp8/1k1k/mtp/ctx1_gen3_tep8_batch2_eplb0_mtp3_9.yaml"
-        decode:
-          num-worker: 3
-          tp: 8
-          ep: 8
-          dp-attn: false
-    # 1k1k STP configs
-      - conc-list: [6144]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp8/1k1k/stp/ctx1_gen1_dep8_batch768_eplb0_mtp0_6144.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp8/1k1k/stp/ctx1_gen1_dep8_batch768_eplb0_mtp0_6144.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [4301]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp8/1k1k/stp/ctx1_gen1_dep8_batch512_eplb0_mtp0_4301.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp8/1k1k/stp/ctx1_gen1_dep8_batch512_eplb0_mtp0_4301.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [2151]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp8/1k1k/stp/ctx1_gen1_dep16_batch128_eplb0_mtp0_2151.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp8/1k1k/stp/ctx1_gen1_dep16_batch128_eplb0_mtp0_2151.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [1127]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp8/1k1k/stp/ctx1_gen1_dep32_batch32_eplb0_mtp0_1127.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp8/1k1k/stp/ctx1_gen1_dep32_batch32_eplb0_mtp0_1127.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [256]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp8/1k1k/stp/ctx1_gen1_dep32_batch8_eplb0_mtp0_256.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp8/1k1k/stp/ctx1_gen1_dep32_batch8_eplb0_mtp0_256.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [27]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp8/1k1k/stp/ctx1_gen3_tep8_batch8_eplb0_mtp0_27.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp8/1k1k/stp/ctx1_gen3_tep8_batch8_eplb0_mtp0_27.yaml"
-        decode:
-          num-worker: 3
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [3]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb200-fp8/1k1k/stp/ctx1_gen3_tep8_batch1_eplb0_mtp0_3.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb200-fp8/1k1k/stp/ctx1_gen3_tep8_batch1_eplb0_mtp0_3.yaml"
-        decode:
-          num-worker: 3
-          tp: 8
-          ep: 8
-          dp-attn: false
-    # 8k1k MTP configs
-
-dsr1-fp8-gb200-dynamo-sglang:
-  image: lmsysorg/sglang:v0.5.8.post1-cu130
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: gb200
-  precision: fp8
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "0.8.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-     # "Low latency" (1 prefill worker at TP4 and 1 decode worker at TP4)
-      - conc-list: [4, 8]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/gb200-fp8/1k1k/low-latency.yaml
-          - "CONFIG_FILE=recipes/gb200-fp8/1k1k/low-latency.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-      # "Mid curve" (3 prefill workers at DEP8 and 1 decode worker at DEP48)
-      - conc-list: [1024, 2048, 4096]
-        prefill:
-          num-worker: 3
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/gb200-fp8/1k1k/mid-curve.yaml
-          - "CONFIG_FILE=recipes/gb200-fp8/1k1k/mid-curve.yaml"
-        decode:
-          num-worker: 1
-          tp: 48
-          ep: 48
-          dp-attn: true
-
-      # "Max throughput" (2 prefill workers at DEP8 and 1 decode worker at DEP32)
-      - conc-list: [1024, 2048, 4096, 6144]
-        prefill:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/gb200-fp8/1k1k/max-tpt.yaml
-          - "CONFIG_FILE=recipes/gb200-fp8/1k1k/max-tpt.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-
-      # "Ultra throughput" (1 prefill workers at DEP8 and 1 decode worker at DEP8)
-      - conc-list: [4096]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/gb200-fp8/1k1k/ultra-tpt.yaml
-          - "CONFIG_FILE=recipes/gb200-fp8/1k1k/ultra-tpt.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-dsr1-fp8-gb300-dynamo-sglang:
-  image: lmsysorg/sglang:v0.5.8.post1-cu130
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: gb300
-  precision: fp8
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "0.8.0" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-     # "Low latency" (1 prefill worker at TP4 and 4 decode workers at TP4)
-      - conc-list: [4, 8, 16, 32]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/gb300-fp8/1k1k/stp/low-latency.yaml
-          - "CONFIG_FILE=recipes/gb300-fp8/1k1k/stp/low-latency.yaml"
-        decode:
-          num-worker: 4
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-      # "Mid curve" (2 prefill workers at DEP8 and 1 decode worker at DEP32)
-      - conc-list: [1024, 2048, 4096, 6144]
-        prefill:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/gb300-fp8/1k1k/stp/mid.yaml
-          - "CONFIG_FILE=recipes/gb300-fp8/1k1k/stp/mid.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-
-      # "Max throughput" (1 prefill worker at DEP8 and 1 decode worker at DEP8)
-      - conc-list: [4096, 7168, 7680]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/gb300-fp8/1k1k/stp/max.yaml
-          - "CONFIG_FILE=recipes/gb300-fp8/1k1k/stp/max.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-dsr1-fp4-gb200-dynamo-sglang:
-  image: "lmsysorg/sglang:v0.5.8-cu130"
-  model: nvidia/DeepSeek-R1-0528-NVFP4-v2
-  model-prefix: dsr1
-  runner: gb200
-  precision: fp4
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "0.8.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # Low latency (1 prefill node, 2 decode nodes)
-      - spec-decoding: "none"
-        conc-list: [ 4, 8, 32 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb200-fp4/1k1k/low-latency.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-      # Mid curve (4 prefill nodes, 8 decode nodes)
-      - spec-decoding: "none"
-        conc-list: [ 512, 2048, 4096, 8192 ]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb200-fp4/1k1k/mid-curve.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-
-      # Max throughput (4 prefill nodes, 12 decode nodes)
-      - spec-decoding: "none"
-        conc-list: [ 2048, 4096 ]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb200-fp4/1k1k/max-tpt.yaml"
-        decode:
-          num-worker: 1
-          tp: 48
-          ep: 48
-          dp-attn: true
-
-    # 8k1k configurations
-
-dsr1-fp4-gb300-dynamo-trt:
-  image: nvcr.io#nvidia/ai-dynamo/tensorrtllm-runtime:0.8.1.post2
-  model: nvidia/DeepSeek-R1-0528-NVFP4-v2
-  model-prefix: dsr1
-  runner: gb300
-  precision: fp4
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "0.8.1.post2" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # MTP configurations
-      - spec-decoding: "mtp"
-        conc-list: [3226]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp4/1k1k/mtp/ctx1_gen1_dep4_batch768_eplb0_mtp1.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp4/1k1k/mtp/ctx1_gen1_dep4_batch768_eplb0_mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [333]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp4/1k1k/mtp/ctx1_gen1_dep32_batch8_eplb0_mtp.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp4/1k1k/mtp/ctx1_gen1_dep32_batch8_eplb0_mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [5]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp4/1k1k/mtp/ctx1_gen4_tep8_batch1_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp4/1k1k/mtp/ctx1_gen4_tep8_batch1_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [8, 12, 24, 48]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp4/1k1k/mtp/ctx1_gen4_tep8_batch8_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp4/1k1k/mtp/ctx1_gen4_tep8_batch8_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [2253]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp4/1k1k/mtp/ctx3_gen1_dep16_batch128_eplb256_mtp1.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp4/1k1k/mtp/ctx3_gen1_dep16_batch128_eplb256_mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [1229]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp4/1k1k/mtp/ctx3_gen1_dep32_batch32_eplb288_mtp3.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp4/1k1k/mtp/ctx3_gen1_dep32_batch32_eplb288_mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      # Non-MTP configurations (default spec_decoding="none")
-      - conc-list: [5]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp4/1k1k/stp/ctx1_gen4_tep8_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp4/1k1k/stp/ctx1_gen4_tep8_batch1_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [12, 48, 96, 192]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp4/1k1k/stp/ctx1_gen4_tep8_batch32_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp4/1k1k/stp/ctx1_gen4_tep8_batch32_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [8192]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp4/1k1k/stp/ctx2_gen1_dep8_batch1024_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp4/1k1k/stp/ctx2_gen1_dep8_batch1024_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [1229]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp4/1k1k/stp/ctx2_gen1_dep32_batch32_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp4/1k1k/stp/ctx2_gen1_dep32_batch32_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [4301]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp4/1k1k/stp/ctx3_gen1_dep16_batch256_eplb256_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp4/1k1k/stp/ctx3_gen1_dep16_batch256_eplb256_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [2253]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp4/1k1k/stp/ctx3_gen1_dep32_batch64_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp4/1k1k/stp/ctx3_gen1_dep32_batch64_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-
-dsr1-fp4-gb300-dynamo-sglang:
-  image: "lmsysorg/sglang:v0.5.8.post1-cu130-runtime"
-  model: nvidia/DeepSeek-R1-0528-NVFP4-v2
-  model-prefix: dsr1
-  runner: gb300
-  precision: fp4
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "0.8.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # Low latency (1 prefill node, 2 decode nodes)
-      - spec-decoding: "none"
-        conc-list: [ 4, 8, 32 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/1k1k/low_latency.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-      # Mid curve (4 prefill nodes, 8 decode nodes)
-      - spec-decoding: "none"
-        conc-list: [ 512, 2048, 4096, 8192 ]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/1k1k/mid_curve.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-
-      # Max throughput (4 prefill nodes, 12 decode nodes)
-      - spec-decoding: "none"
-        conc-list: [ 512, 2048, 4096, 8192 ]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/1k1k/max_tpt.yaml"
-        decode:
-          num-worker: 1
-          tp: 48
-          ep: 48
-          dp-attn: true
-
-    # 8k1k configurations
-
-dsr1-fp8-gb300-dynamo-trt:
-  image: nvcr.io/nvidia/ai-dynamo/tensorrtllm-runtime:1.3.0-dev.1-cuda13
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: gb300-nv
-  precision: fp8
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "v1.3.0-dev.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # MTP configurations (spec_decoding="mtp")
-      - spec-decoding: "mtp"
-        conc-list: [8]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp8/1k1k/mtp/ctx1_gen4_tep8_batch1_eplb0_mtp3_8.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp8/1k1k/mtp/ctx1_gen4_tep8_batch1_eplb0_mtp3_8.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [24]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp8/1k1k/mtp/ctx1_gen4_tep8_batch4_eplb0_mtp3_24.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp8/1k1k/mtp/ctx1_gen4_tep8_batch4_eplb0_mtp3_24.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [180]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp8/1k1k/mtp/ctx1_gen1_dep32_batch4_eplb0_mtp3_180.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp8/1k1k/mtp/ctx1_gen1_dep32_batch4_eplb0_mtp3_180.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [564]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp8/1k1k/mtp/ctx2_gen1_dep32_batch16_eplb0_mtp3_564.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp8/1k1k/mtp/ctx2_gen1_dep32_batch16_eplb0_mtp3_564.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [666]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp8/1k1k/mtp/ctx1_gen1_dep16_batch32_eplb0_mtp3_666.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp8/1k1k/mtp/ctx1_gen1_dep16_batch32_eplb0_mtp3_666.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [2253]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp8/1k1k/mtp/ctx2_gen1_dep16_batch128_eplb0_mtp1_2253.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp8/1k1k/mtp/ctx2_gen1_dep16_batch128_eplb0_mtp1_2253.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [8192]
-        prefill:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp8/1k1k/mtp/ctx3_gen2_dep8_batch512_eplb0_mtp1_8192.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp8/1k1k/mtp/ctx3_gen2_dep8_batch512_eplb0_mtp1_8192.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # STP configurations (no spec_decoding)
-      - conc-list: [4]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp8/1k1k/stp/ctx1_gen4_tep8_batch1_eplb0_mtp0_4.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp8/1k1k/stp/ctx1_gen4_tep8_batch1_eplb0_mtp0_4.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [24]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp8/1k1k/stp/ctx1_gen4_tep8_batch4_eplb0_mtp0_24.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp8/1k1k/stp/ctx1_gen4_tep8_batch4_eplb0_mtp0_24.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [84]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp8/1k1k/stp/ctx1_gen4_tep8_batch16_eplb0_mtp0_84.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp8/1k1k/stp/ctx1_gen4_tep8_batch16_eplb0_mtp0_84.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [1229]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp8/1k1k/stp/ctx2_gen1_dep32_batch32_eplb0_mtp0_1229.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp8/1k1k/stp/ctx2_gen1_dep32_batch32_eplb0_mtp0_1229.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [2253]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp8/1k1k/stp/ctx2_gen1_dep16_batch128_eplb0_mtp0_2253.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp8/1k1k/stp/ctx2_gen1_dep16_batch128_eplb0_mtp0_2253.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [8602]
-        prefill:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp8/1k1k/stp/ctx3_gen2_dep8_batch512_eplb0_mtp0_8602.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp8/1k1k/stp/ctx3_gen2_dep8_batch512_eplb0_mtp0_8602.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [12288]
-        prefill:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/trtllm/gb300-fp8/1k1k/stp/ctx3_gen2_dep8_batch768_eplb0_mtp0_12288.yaml
-          - "CONFIG_FILE=recipes/trtllm/gb300-fp8/1k1k/stp/ctx3_gen2_dep8_batch768_eplb0_mtp0_12288.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-dsr1-fp8-h200-dynamo-sglang:
-  image: lmsysorg/sglang:v0.5.8.post1-cu130
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: cluster:h200-dgxc
-  precision: fp8
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "0.8.0" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # STP: Low latency (1 prefill, 9 decode, TEP)
-      - spec-decoding: "none"
-        conc-list: [1, 4, 8, 16, 32, 64, 128, 256]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/h200/1k1k/low-latency-1p9d.yaml"
-        decode:
-          num-worker: 9
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # STP: High throughput TEP (1 prefill, 6 decode)
-      - spec-decoding: "none"
-        conc-list: [512, 1024, 2048]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/h200/1k1k/bs256-1p6d-tp.yaml"
-        decode:
-          num-worker: 6
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # STP: High throughput DEP (1 prefill, 6 decode, dp-attention)
-      - spec-decoding: "none"
-        conc-list: [128, 256, 512, 1024, 2048]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/h200/1k1k/bs256-1p6d-dep.yaml"
-        decode:
-          num-worker: 6
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # MTP: Low latency (1 prefill, 9 decode, TEP)
-      - spec-decoding: "mtp"
-        conc-list: [1, 4, 8, 16, 32, 64, 128, 256]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/h200/1k1k/low-latency-1p9d-mtp.yaml"
-        decode:
-          num-worker: 9
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # MTP: High throughput TEP (1 prefill, 6 decode)
-      - spec-decoding: "mtp"
-        conc-list: [512, 1024, 2048]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/h200/1k1k/bs256-1p6d-tp-mtp.yaml"
-        decode:
-          num-worker: 6
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # MTP: High throughput DEP (1 prefill, 6 decode, dp-attention)
-      - spec-decoding: "mtp"
-        conc-list: [128, 256, 512, 1024, 2048]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/h200/1k1k/bs256-1p6d-dep-mtp.yaml"
-        decode:
-          num-worker: 6
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-dsr1-fp4-b200-dynamo-sglang:
-  image: lmsysorg/sglang:v0.5.8.post1-cu130-runtime
-  model: deepseek-r1-fp4
-  model-prefix: dsr1
-  runner: cluster:b200-dgxc
-  precision: fp4
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "0.8.0" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # Non-MTP configurations
-      - conc-list: [16, 128]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp4/1k1k.yaml:zip_override_stp_lowlat[0]"
-        decode:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [32, 64, 256]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp4/1k1k.yaml:zip_override_stp_lowlat[1]"
-        decode:
-          num-worker: 6
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [512]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp4/1k1k.yaml:zip_override_stp_maxtpt[0]"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [512]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp4/1k1k.yaml:zip_override_stp_maxtpt[1]"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-dsr1-fp8-b200-dynamo-sglang:
-  image: lmsysorg/sglang:v0.5.8.post1-cu130-amd64
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: cluster:b200-dgxc
-  precision: fp8
-  framework: dynamo-sglang
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # Non-MTP configurations
-      - conc-list: [4]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/1k1k.yaml:zip_override_stp_lowlat[0]"
-        router: { name: dynamo-router, version: "0.8.0" }
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [16, 32, 64, 128, 256]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/1k1k.yaml:zip_override_stp_lowlat[1]"
-        router: { name: dynamo-router, version: "0.8.0" }
-        decode:
-          num-worker: 3
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [1024, 2048, 4096]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/1k1k.yaml:zip_override_stp_maxtpt[0]"
-        router: { name: dynamo-router, version: "0.8.0" }
-        decode:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [2048, 4096]
-        prefill:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/1k1k.yaml:zip_override_stp_maxtpt[1]"
-        router: { name: dynamo-router, version: "0.8.0" }
-        decode:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-dsr1-fp8-b200-dynamo-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.8.post1-cu130-amd64
-  model: deepseek-ai/DeepSeek-R1-0528
-  model-prefix: dsr1
-  runner: cluster:b200-dgxc
-  precision: fp8
-  framework: dynamo-sglang
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # MTP low-latency: 1P1D
-      - spec-decoding: "mtp"
-        conc-list: [4, 64]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/1k1k.yaml:zip_override_mtp_lowlat[0]"
-        router: { name: dynamo-router, version: "0.8.0" }
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: false
-      # MTP low-latency: 1P3D
-      - spec-decoding: "mtp"
-        conc-list: [4, 8, 16, 32, 128]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/1k1k.yaml:zip_override_mtp_lowlat[1]"
-        router: { name: dynamo-router, version: "0.8.0" }
-        decode:
-          num-worker: 3
-          tp: 8
-          ep: 8
-          dp-attn: false
-      # MTP max-tpt: 1P5D
-      - spec-decoding: "mtp"
-        conc-list: [512, 4096]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/1k1k.yaml:zip_override_mtp_maxtpt[1]"
-        router: { name: dynamo-router, version: "0.8.0" }
-        decode:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # MTP max-tpt: 2P5D
-      - spec-decoding: "mtp"
-        conc-list: [1024, 2048, 4096]
-        prefill:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/1k1k.yaml:zip_override_mtp_maxtpt[2]"
-        router: { name: dynamo-router, version: "0.8.0" }
-        decode:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # MTP max-tpt: 1P2D
-      - spec-decoding: "mtp"
-        conc-list: [512, 1024, 2048]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/1k1k.yaml:override_mtp_maxtpt_1p2d"
-        router: { name: dynamo-router, version: "0.8.0" }
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-dsr1-fp4-b200-dynamo-sglang-mtp:
-  image: "lmsysorg/sglang:v0.5.12.post1"
-  model: deepseek-r1-fp4
-  model-prefix: dsr1
-  runner: cluster:b200-dgxc
-  precision: fp4
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "5b4bc1dd70965017a737c71b19db5a0aeaa88727" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - spec-decoding: "mtp"
-        conc-list: [16, 512]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsr1/sglang/b200-fp4/1k1k/disagg-1p5d-p-tp4-ep4-dp4-d-tp8-ep8-b512-c16x512-mtp.yaml"
-        decode:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [32, 64, 256, 512]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsr1/sglang/b200-fp4/1k1k/disagg-1p6d-p-tp4-ep4-dp4-d-tp8-ep8-b512-c32x64x256x512-mtp.yaml"
-        decode:
-          num-worker: 6
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [512, 1024]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsr1/sglang/b200-fp4/1k1k/disagg-1p1d-p-tp4-ep4-dp4-d-tp8-ep8-dp8-b1024-c512x1024-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [512]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsr1/sglang/b200-fp4/1k1k/disagg-1p2d-p-tp4-ep4-dp4-d-tp8-ep8-dp8-b512-c512-mtp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-# Original config key: kimik2.5-fp4-gb200-dynamo-trt
-kimik2.5-fp4-gb200-dynamo-trt-deprecated-1k1k:
-  image: nvcr.io#nvidia/ai-dynamo/tensorrtllm-runtime:1.3.0-dev.1-cuda13
-  model: nvidia/Kimi-K2.5-NVFP4
-  model-prefix: kimik2.5
-  runner: gb200
-  precision: fp4
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "v1.3.0-dev.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # Non-MTP configurations (default spec_decoding="none")
-      - conc-list: [ 8 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch1_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 12 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch2_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch2_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 24 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch4_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch4_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 192 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch32_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch32_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 30 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen5tep4_batch4_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen5tep4_batch4_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 60 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen5tep4_batch8_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen5tep4_batch8_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 333 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch8_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch8_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 666 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch16_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch16_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 1127 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch32_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch32_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 4301 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep16_batch256_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep16_batch256_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [ 8192 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep16_batch512_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep16_batch512_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [ 2253 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep32_batch64_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep32_batch64_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 4301 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep32_batch128_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep32_batch128_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-
-# Original config key: kimik2.5-fp4-gb300-dynamo-trt
-kimik2.5-fp4-gb300-dynamo-trt-deprecated-1k1k:
-  image: nvcr.io#nvidia/ai-dynamo/tensorrtllm-runtime:1.3.0-dev.1-cuda13
-  model: nvidia/Kimi-K2.5-NVFP4
-  model-prefix: kimik2.5
-  runner: gb300
-  precision: fp4
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "v1.3.0-dev.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # Non-MTP configurations (default spec_decoding="none")
-      - conc-list: [ 8 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch1_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 12 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch2_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch2_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 24 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch4_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch4_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 30 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen5tep4_batch4_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen5tep4_batch4_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 60 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen5tep4_batch8_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen5tep4_batch8_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 333 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch8_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch8_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 666 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch16_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch16_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 1229 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch32_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch32_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 4301 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep16_batch256_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep16_batch256_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [ 8192 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep16_batch512_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep16_batch512_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [ 2253 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep32_batch64_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep32_batch64_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 4301 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep32_batch128_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep32_batch128_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-
-# Original config key: kimik2.5-fp4-gb200-dynamo-vllm
-kimik2.5-fp4-gb200-dynamo-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai:v0.21.0
-  model: nvidia/Kimi-K2.5-NVFP4
-  model-prefix: kimik2.5
-  runner: gb200
-  precision: fp4
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.2.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - conc-list: [4096, 12288]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.5/vllm/gb200-fp4/1k1k/disagg-1p1d-dep4-dep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 1
-          ep: 8
-          dp-attn: true
-      - conc-list: [4, 8, 32, 128]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.5/vllm/gb200-fp4/1k1k/disagg-1p4d-dep4-tp8-stp.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [4096, 6144]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.5/vllm/gb200-fp4/1k1k/disagg-1p1d-dep4-dep16-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 1
-          ep: 16
-          dp-attn: true
-
-qwen3.5-fp8-gb200-dynamo-sglang:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260608-303757cc
-  model: Qwen/Qwen3.5-397B-A17B-FP8
-  model-prefix: qwen3.5
-  runner: gb200
-  precision: fp8
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "46520ca59afe992fb5ef61b3197b2316f8df9b2b" }
-  kv-p2p-transfer: mooncake
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 1P1D STP: TP4 prefill + TP4 decode (pure tensor parallel). 2 nodes (1+1).
-      - spec-decoding: "none"
-        conc-list: [1, 2, 4, 8, 16, 32, 64]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/qwen3.5/sglang/gb200-fp8/1k1k/disagg-1p1d-tp4-tp4-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1P1D wide-EP: prefill DEP4 + decode DEP16. 5 nodes (1+4).
-      - spec-decoding: "none"
-        conc-list: [512, 1024, 2048]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/qwen3.5/sglang/gb200-fp8/1k1k/disagg-1p1d-dep4-dep16-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # 2P1D wide-EP: 2 prefill DEP4 + decode DEP16. 6 nodes (2+4).
-      - spec-decoding: "none"
-        conc-list: [4096]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/qwen3.5/sglang/gb200-fp8/1k1k/disagg-2p1d-dep4-dep16-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-# Original config key: glm5-fp4-gb200-dynamo-sglang-mtp
-glm5-fp4-gb200-dynamo-sglang-mtp-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.13.post1-cu130
-  model: nvidia/GLM-5.1-NVFP4
-  model-prefix: glm5.1
-  runner: gb200
-  precision: fp4
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "1.2.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 1p1d dep16. 5 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [1742]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_1k1k_hightpt[0]"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # 1p1d dep16, conc1024. 5 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [1150]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_1k1k_hightpt[1]"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # 1p1d dep16, conc256. 5 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [302]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_1k1k_hightpt[2]"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # 1p1d dep32. 9 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [1263]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_1k1k_hightpt[3]"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-    # ---------- 1k1k low-latency (per-node TP=4 decode workers, MTP) ----------
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 1p16d tp4, conc1. 17 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [26]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_1k1k_lowlat[0]"
-        decode:
-          num-worker: 16
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p16d tp4, conc16. 17 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [343]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_1k1k_lowlat[1]"
-        decode:
-          num-worker: 16
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p16d tp4, conc4. 17 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [94]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_1k1k_lowlat[2]"
-        decode:
-          num-worker: 16
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-# Original config key: dsv4-fp4-gb300-dynamo-trt
-dsv4-fp4-gb300-dynamo-trt-deprecated-1k1k:
-  image: nvcr.io#nvidia/ai-dynamo/tensorrtllm-runtime:1.3.0-deepseek-v4-dev.1
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: gb300-nv
-  precision: fp4
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "v1.3.0-deepseek-v4-dev.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - conc-list: [4]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx1dep4_gen4tep8_batch1_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [5]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx1dep4_gen5tep4_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx1dep4_gen5tep4_batch1_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [15]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx1dep4_gen5tep4_batch2_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx1dep4_gen5tep4_batch2_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [25]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx1dep4_gen5tep4_batch4_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx1dep4_gen5tep4_batch4_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [55]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx1dep4_gen5tep4_batch8_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx1dep4_gen5tep4_batch8_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [167]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch4_eplb384_mtp0.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch4_eplb384_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [333]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch8_eplb384_mtp0.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch8_eplb384_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [666]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch16_eplb384_mtp0.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch16_eplb384_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [1229]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch32_eplb384_mtp0.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx1dep4_gen1dep32_batch32_eplb384_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [2253]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep32_batch64_eplb384_mtp0.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx2dep4_gen1dep32_batch64_eplb384_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [4301]
-        prefill:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx3dep4_gen1dep32_batch128_eplb384_mtp0.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx3dep4_gen1dep32_batch128_eplb384_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [8192]
-        prefill:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx3dep4_gen1dep16_batch512_eplb384_mtp0.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx3dep4_gen1dep16_batch512_eplb384_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [8192]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx4dep4_gen1dep32_batch256_eplb384_mtp0.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx4dep4_gen1dep32_batch256_eplb384_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [16384]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx4dep4_gen1dep16_batch1024_eplb384_mtp0.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/STP/ctx4dep4_gen1dep16_batch1024_eplb384_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-# Original config key: dsv4-fp4-gb300-dynamo-trt-mtp
-dsv4-fp4-gb300-dynamo-trt-mtp-deprecated-1k1k:
-  image: nvcr.io#nvidia/ai-dynamo/tensorrtllm-runtime:1.3.0-deepseek-v4-dev.1
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: gb300-nv
-  precision: fp4
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "v1.3.0-deepseek-v4-dev.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - spec-decoding: "mtp"
-        conc-list: [8]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen4tep8_batch1_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen4tep8_batch1_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [10]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen5tep4_batch1_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen5tep4_batch1_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [15]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen5tep4_batch2_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen5tep4_batch2_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [25]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen5tep4_batch4_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen5tep4_batch4_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [90]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen1dep32_batch2_eplb384_mtp3.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen1dep32_batch2_eplb384_mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [167]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen1dep32_batch4_eplb384_mtp3.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen1dep32_batch4_eplb384_mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [333]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen1dep32_batch8_eplb384_mtp3.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen1dep32_batch8_eplb384_mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [615]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen1dep32_batch16_eplb384_mtp3.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx1dep4_gen1dep32_batch16_eplb384_mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [1229]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx2dep4_gen1dep32_batch32_eplb384_mtp3.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx2dep4_gen1dep32_batch32_eplb384_mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [2253]
-        prefill:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx3dep4_gen1dep32_batch64_eplb384_mtp3.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx3dep4_gen1dep32_batch64_eplb384_mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [4301]
-        prefill:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx3dep4_gen1dep16_batch256_eplb384_mtp1.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx3dep4_gen1dep16_batch256_eplb384_mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [4301]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx4dep4_gen1dep32_batch128_eplb384_mtp1.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx4dep4_gen1dep32_batch128_eplb384_mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [8192]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx2dep4_gen1dep8_batch1024_eplb384_mtp1.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx2dep4_gen1dep8_batch1024_eplb384_mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [8192]
-        prefill:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx3dep4_gen1dep16_batch512_eplb384_mtp1.yaml
-          - "CONFIG_FILE=recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL1K_OSL1K/MTP/ctx3dep4_gen1dep16_batch512_eplb384_mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-# Original config key: glm5-fp8-b200-dynamo-sglang
-glm5-fp8-b200-dynamo-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.11-cu130
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: b200-dgxc
-  precision: fp8
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "0.8.0" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - conc-list: [2576]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/glm5.yaml:zip_override_1k1k_hightpt[0]"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-      - conc-list: [1248]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/glm5.yaml:zip_override_1k1k_hightpt[1]"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: true
-      - conc-list: [800]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/glm5.yaml:zip_override_1k1k_hightpt[2]"
-        decode:
-          num-worker: 3
-          tp: 8
-          ep: 1
-          dp-attn: true
-      - conc-list: [576]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/glm5.yaml:zip_override_1k1k_hightpt[3]"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 1
-          dp-attn: true
-      - conc-list: [512, 256, 128, 64, 32]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/glm5.yaml:zip_override_1k1k_lowlat[0]"
-        decode:
-          num-worker: 8
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [16]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/glm5.yaml:zip_override_1k1k_lowlat[1]"
-        decode:
-          num-worker: 8
-          tp: 8
-          ep: 1
-          dp-attn: false
-
-qwen3.5-fp8-h100-sglang:
-  image: lmsysorg/sglang:v0.5.14-cu130
-  model: Qwen/Qwen3.5-397B-A17B-FP8
-  model-prefix: qwen3.5
-  runner: h100
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 8 }
-      - { tp: 8, ep: 8, conc-start: 16, conc-end: 256 }
-
-qwen3.5-fp8-h100-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.14-cu130
-  model: Qwen/Qwen3.5-397B-A17B-FP8
-  model-prefix: qwen3.5
-  runner: h100
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 8, conc-start: 4, conc-end: 32, spec-decoding: mtp }
-
-# Original config key: glm5-fp4-gb200-dynamo-sglang
-glm5-fp4-gb200-dynamo-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.11-cu130
-  model: nvidia/GLM-5.1-NVFP4
-  model-prefix: glm5.1
-  runner: gb200
-  precision: fp4
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "1.2.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 2p1d wide-EP. 10 nodes.
-      - conc-list: [3160]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/1k1k/disagg-2p1d-p-tp4-ep4-dp4-d-tp32-ep32-dp32-b2048-c3160-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-    # ---------- 1k1k low-latency ----------
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 1p1d wide-EP dep16. 5 nodes.
-      - conc-list: [2115]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/1k1k/disagg-1p1d-p-tp4-ep4-dp4-d-tp16-ep16-dp16-b2023-c2115-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # 1p1d wide-EP dep32, mrr1024. 9 nodes.
-      - conc-list: [1156]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/1k1k/disagg-1p1d-p-tp4-ep4-dp4-d-tp32-ep32-dp32-b1024-c1156-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      # 1p1d wide-EP dep32, mrr512. 9 nodes.
-      - conc-list: [556]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/1k1k/disagg-1p1d-p-tp4-ep4-dp4-d-tp32-ep32-dp32-b512-c556-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      # 1p16d, mrr1 (single-stream). 17 nodes.
-      - conc-list: [23]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/1k1k/disagg-1p16d-p-tp4-ep4-dp4-d-tp4-b1-c23-stp.yaml"
-        decode:
-          num-worker: 16
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p16d, mrr16. 17 nodes.
-      - conc-list: [290]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/1k1k/disagg-1p16d-p-tp4-ep4-dp4-d-tp4-b16-c290-stp.yaml"
-        decode:
-          num-worker: 16
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p16d, mrr4. 17 nodes.
-      - conc-list: [73]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/1k1k/disagg-1p16d-p-tp4-ep4-dp4-d-tp4-b4-c73-stp.yaml"
-        decode:
-          num-worker: 16
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-# Original config key: glm5-fp4-gb300-dynamo-sglang
-glm5-fp4-gb300-dynamo-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.11-cu130
-  model: nvidia/GLM-5.1-NVFP4
-  model-prefix: glm5
-  runner: gb300-nv
-  precision: fp4
-  framework: dynamo-sglang
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 3p1d wide-EP. 11 nodes. conc 16500.
-      - conc-list: [16500]
-        prefill:
-          num-worker: 3
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5.yaml:zip_override_1k1k_hightpt[0]"
-        router: { name: dynamo-router, version: "0.8.0" }
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      # 2p1d wide-EP. 10 nodes. conc 8300.
-      - conc-list: [8300]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5.yaml:zip_override_1k1k_hightpt[1]"
-        router: { name: dynamo-router, version: "0.8.0" }
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      # 1p1d wide-EP. 9 nodes. conc sweep 2500x1024x512x256.
-      - conc-list: [2500, 1024, 512, 256]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5.yaml:zip_override_1k1k_hightpt[2]"
-        router: { name: dynamo-router, version: "0.8.0" }
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-    # ---------- 1k1k low-latency (per-node TP=4 decode workers) ----------
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 1p17d low-latency, bs=32 sweep. 18 nodes.
-      - conc-list: [512, 256, 128, 64]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5.yaml:zip_override_1k1k_lowlat[0]"
-        router: { name: dynamo-router, version: "0.8.0" }
-        decode:
-          num-worker: 17
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p17d low-latency, bs=1 (single-stream). 18 nodes.
-      - conc-list: [32]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5.yaml:zip_override_1k1k_lowlat[1]"
-        router: { name: dynamo-router, version: "0.8.0" }
-        decode:
-          num-worker: 17
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-# Original config key: glm5-fp8-gb200-dynamo-sglang
-glm5-fp8-gb200-dynamo-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.12
-  model: zai-org/GLM-5.1-FP8
-  model-prefix: glm5.1
-  runner: gb200
-  precision: fp8
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "1.2.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 2p1d wide-EP (dep32, mrr1024). 12 nodes.
-      - conc-list: [2161]
-        prefill:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/1k1k/disagg-2p1d-p-tp8-dp8-d-tp32-ep32-dp32-b1024-c2161-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-    # ---------- 1k1k low-latency ----------
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 1p1d wide-EP (dep16). 6 nodes.
-      - conc-list: [1955]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/1k1k/disagg-1p1d-p-tp8-dp8-d-tp16-ep16-dp16-b1815-c1955-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # 1p1d wide-EP (dep16, mrr1024). 6 nodes.
-      - conc-list: [1170]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/1k1k/disagg-1p1d-p-tp8-dp8-d-tp16-ep16-dp16-b1024-c1170-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # 1p1d wide-EP (dep16, mrr256). 6 nodes.
-      - conc-list: [298]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/1k1k/disagg-1p1d-p-tp8-dp8-d-tp16-ep16-dp16-b256-c298-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # 1p6d per-node TP=8, mrr1 (single-stream). 14 nodes.
-      - conc-list: [8]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/1k1k/disagg-1p6d-p-tp8-dp8-d-tp8-b1-c8-stp.yaml"
-        decode:
-          num-worker: 6
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # 1p6d per-node TP=8, mrr16. 14 nodes.
-      - conc-list: [72]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/1k1k/disagg-1p6d-p-tp8-dp8-d-tp8-b16-c72-stp.yaml"
-        decode:
-          num-worker: 6
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # 1p6d per-node TP=8, mrr4. 14 nodes.
-      - conc-list: [20]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/1k1k/disagg-1p6d-p-tp8-dp8-d-tp8-b4-c20-stp.yaml"
-        decode:
-          num-worker: 6
-          tp: 8
-          ep: 1
-          dp-attn: false
-
-# Original config key: glm5-fp4-gb300-dynamo-sglang-mtp
-glm5-fp4-gb300-dynamo-sglang-mtp-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.13.post1-cu130
-  model: nvidia/GLM-5.1-NVFP4
-  model-prefix: glm5.1
-  runner: gb300-nv
-  precision: fp4
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "0.8.0" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 1p1d dep16. 5 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [773]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5-mtp.yaml:zip_override_mtp_1k1k_hightpt[0]"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # 1p1d dep16, conc512. 5 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [578]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5-mtp.yaml:zip_override_mtp_1k1k_hightpt[1]"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # 1p1d dep32. 9 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [677]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5-mtp.yaml:zip_override_mtp_1k1k_hightpt[2]"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      # 2p1d dep16. 5 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [2048]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5-mtp.yaml:zip_override_mtp_1k1k_hightpt[3]"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # 2p1d dep16, conc1024. 5 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [1362]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5-mtp.yaml:zip_override_mtp_1k1k_hightpt[4]"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-    # ---------- 1k1k low-latency (per-node TP=4 decode workers, MTP) ----------
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 1p8d tp4, conc1. 9 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [13]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5-mtp.yaml:zip_override_mtp_1k1k_lowlat[0]"
-        decode:
-          num-worker: 8
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p8d tp4, conc16. 9 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [162]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5-mtp.yaml:zip_override_mtp_1k1k_lowlat[1]"
-        decode:
-          num-worker: 8
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p8d tp4, conc4. 9 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [44]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5-mtp.yaml:zip_override_mtp_1k1k_lowlat[2]"
-        decode:
-          num-worker: 8
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-# Original config key: glm5-fp8-gb300-dynamo-sglang
-glm5-fp8-gb300-dynamo-sglang-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.11-cu130
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: gb300-nv
-  precision: fp8
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "0.8.0" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - conc-list: [8192]
-        prefill:
-          num-worker: 12
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp8/glm5.yaml:zip_override_1k1k_hightpt[0]"
-        decode:
-          num-worker: 1
-          tp: 24
-          ep: 24
-          dp-attn: true
-      - conc-list: [7500]
-        prefill:
-          num-worker: 10
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp8/glm5.yaml:zip_override_1k1k_hightpt[1]"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [7300]
-        prefill:
-          num-worker: 8
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp8/glm5.yaml:zip_override_1k1k_hightpt[2]"
-        decode:
-          num-worker: 1
-          tp: 40
-          ep: 40
-          dp-attn: true
-      - conc-list: [6500]
-        prefill:
-          num-worker: 6
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp8/glm5.yaml:zip_override_1k1k_hightpt[3]"
-        decode:
-          num-worker: 1
-          tp: 48
-          ep: 48
-          dp-attn: true
-      - conc-list: [5700]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp8/glm5.yaml:zip_override_1k1k_hightpt[4]"
-        decode:
-          num-worker: 1
-          tp: 56
-          ep: 56
-          dp-attn: true
-    # ---------- 1k1k low-latency (per-node TP=4 decode workers) ----------
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - conc-list: [512, 256, 128, 64]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp8/glm5.yaml:zip_override_1k1k_lowlat[0]"
-        decode:
-          num-worker: 17
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [32]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp8/glm5.yaml:zip_override_1k1k_lowlat[1]"
-        decode:
-          num-worker: 17
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-# Original config key: glm5-fp8-gb300-dynamo-sglang-mtp
-glm5-fp8-gb300-dynamo-sglang-mtp-deprecated-1k1k:
-  image: lmsysorg/sglang:v0.5.11-cu130
-  model: zai-org/GLM-5.1-FP8
-  model-prefix: glm5.1
-  runner: gb300-nv
-  precision: fp8
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "1.2.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - spec-decoding: "mtp"
-        conc-list: [6500]
-        prefill:
-          num-worker: 6
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb300-fp8/1k1k/disagg-6p1d-p-tp4-dp4-d-tp48-ep48-dp48-b6144-c6500-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 48
-          ep: 48
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [5700]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb300-fp8/1k1k/disagg-4p1d-p-tp4-dp4-d-tp56-ep56-dp56-b5600-c5700-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 56
-          ep: 56
-          dp-attn: true
-    # ---------- 1k1k low-latency (per-node TP=4 decode workers, EAGLE MTP) ----------
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - spec-decoding: "mtp"
-        conc-list: [512, 256, 128, 64]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb300-fp8/1k1k/disagg-1p17d-p-tp4-dp4-d-tp4-b32-c512x256x128x64-mtp.yaml"
-        decode:
-          num-worker: 17
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [32]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb300-fp8/1k1k/disagg-1p17d-p-tp4-dp4-d-tp4-b1-c32-mtp.yaml"
-        decode:
-          num-worker: 17
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-# ============================================================================
-# Net-new agentic recipes from chore/agentx-v0.3 (no overlap with main entries).
-# Recipes that ALREADY existed on main were intentionally left at main's version
-# to preserve main behavior; PR-branch modifications to those recipes are NOT
-# brought in here.
-# ============================================================================
-
-# Original config key: minimaxm3-fp8-b300-dynamo-vllm
-minimaxm3-fp8-b300-dynamo-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai:minimax-m3-0618-x86_64-cu130
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: b300
-  precision: fp8
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.3.0.dev20260614" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - conc-list: [4, 16, 64, 128, 4096]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp8/1k1k/disagg-1p1d-dep2-tep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [1, 4, 8, 16]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp8/1k1k/disagg-1p1d-dep2-tp4-marlin-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [2048]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp8/1k1k/disagg-1p2d-dep2-dep4-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-      - conc-list: [512, 4096]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp8/1k1k/disagg-2p1d-dep2-dep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [32]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp8/1k1k/disagg-2p1d-dep2-tep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [16]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp8/1k1k/disagg-2p2d-dep2-tep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [4]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp8/1k1k/disagg-3p2d-dep2-tep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: false
-
-minimaxm3-fp4-b300-dynamo-vllm:
-  image: vllm/vllm-openai:vllm-minimax-m3-perf-x86_64-13.0.1-8b00f41
-  model: nvidia/MiniMax-M3-NVFP4
-  model-prefix: minimaxm3
-  runner: b300
-  precision: fp4
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.3.0.dev20260614" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - conc-list: [4, 16, 64, 128, 4096]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/1k1k/disagg-1p1d-dep2-tep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [1, 4, 8, 16]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/1k1k/disagg-1p1d-dep2-tp4-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [2048]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/1k1k/disagg-1p2d-dep2-dep4-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-      - conc-list: [512, 4096]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/1k1k/disagg-2p1d-dep2-dep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [32]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/1k1k/disagg-2p1d-dep2-tep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [16]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/1k1k/disagg-2p2d-dep2-tep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [4]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/1k1k/disagg-3p2d-dep2-tep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: false
-
-# MiniMax-M3 NVFP4 B300 8k1k disagg sweep with TP1 prefill. This is separate
-# from the legacy entry above so updating the 8k1k frontier does not rerun 1k1k.
-# Original config key: minimaxm3-fp8-gb300-dynamo-vllm
-minimaxm3-fp8-gb300-dynamo-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai:minimax-m3-perf-arm64-13.0.1-7a67223
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: gb300-nv
-  precision: fp8
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.3.0.dev20260614" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 1p1d DEP2+DEP4, 2n: conc 8192
-      - conc-list: [8192]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/1k1k/disagg-1p1d-dep2-dep4-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-
-      # 1p2d DEP2+TEP8, 5n: conc 4,16,64,128,256
-      - conc-list: [4, 16, 64, 128, 256]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/1k1k/disagg-1p2d-dep2-tep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: false
-
-      # 2p2d DEP2+TEP8, 5n: conc 32
-      - conc-list: [32]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/1k1k/disagg-2p2d-dep2-tep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: false
-
-      # 2p3d DEP2+DEP4, 4n: conc 8192
-      - conc-list: [8192]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/1k1k/disagg-2p3d-dep2-dep4-stp.yaml"
-        decode:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-
-      # 2p4d DEP2+DEP4, 5n: conc 8192
-      - conc-list: [8192]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/1k1k/disagg-2p4d-dep2-dep4-stp.yaml"
-        decode:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-
-      # 4p2d DEP2+DEP8, 6n: conc 1024,4096
-      - conc-list: [1024, 4096]
-        prefill:
-          num-worker: 4
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/1k1k/disagg-4p2d-dep2-dep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-qwen3.5-fp4-b200-trt:
-  image: nvcr.io#nvidia/tensorrt-llm/release:1.3.0rc18
-  model: nvidia/Qwen3.5-397B-A17B-NVFP4
-  model-prefix: qwen3.5
-  runner: b200
-  precision: fp4
-  framework: trt
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, ep: 1, conc-list: [8, 16] }
-      - { tp: 4, ep: 4, conc-list: [64, 128] }
-      - { tp: 8, ep: 8, conc-list: [4, 64] }
-      - { tp: 4, ep: 4, dp-attn: true, conc-list: [1024] }
-      - { tp: 8, ep: 8, dp-attn: true, conc-list: [512, 1024] }
-
-# Original config key: minimaxm3-fp8-gb200-dynamo-vllm
-minimaxm3-fp8-gb200-dynamo-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai:minimax-m3-perf-arm64-13.0.1-7a67223
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: gb200
-  precision: fp8
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.3.0.dev20260614" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 1p1d DEP4+DEP8, 3n: conc 1024,4096
-      - conc-list: [1024, 4096]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb200-fp8/1k1k/disagg-1p1d-dep4-dep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-      # 1p1d DEP4+TEP8, 3n: conc 128,256
-      - conc-list: [128, 256]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb200-fp8/1k1k/disagg-1p1d-dep4-tep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: false
-
-      # 1p2d DEP4+DEP8, 5n: conc 1024,4096
-      - conc-list: [1024, 4096]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb200-fp8/1k1k/disagg-1p2d-dep4-dep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-      # 1p2d DEP4+TEP8, 5n: conc 4,16,64
-      - conc-list: [4, 16, 64]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb200-fp8/1k1k/disagg-1p2d-dep4-tep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: false
-
-      # 1p4d DEP4+TP4 Marlin, 5n: conc 32
-      - conc-list: [32]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb200-fp8/1k1k/disagg-1p4d-dep4-tp4-marlin-stp.yaml"
-        decode:
-          num-worker: 4
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-qwen3.5-fp4-b200-trt-mtp:
-  image: nvcr.io#nvidia/tensorrt-llm/release:1.3.0rc18
-  model: nvidia/Qwen3.5-397B-A17B-NVFP4
-  model-prefix: qwen3.5
-  runner: b200
-  precision: fp4
-  framework: trt
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 2, ep: 1, spec-decoding: "mtp", conc-list: [8] }
-      - { tp: 2, ep: 2, spec-decoding: "mtp", conc-list: [4] }
-      - { tp: 8, ep: 8, spec-decoding: "mtp", conc-list: [4] }
-      - { tp: 8, ep: 8, dp-attn: true, spec-decoding: "mtp", conc-list: [64, 128, 256, 512, 1024] }
-
-# Original config key: minimaxm3-fp8-b300-vllm
-minimaxm3-fp8-b300-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai:minimax-m3-0618-x86_64-cu130
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: b300
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 64 }
-      - { tp: 8, ep: 8, conc-start: 1, conc-end: 512 }
-      - { tp: 4, conc-start: 1, conc-end: 64 }
-      - { tp: 4, ep: 4, conc-start: 64, conc-end: 512 }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 128, conc-end: 512 }
-      # tp2 fits MXFP8 weights (~222 GB/GPU of 288) but KV headroom is thin;
-      # 1k1k only, drop if it OOMs at the high end.
-      - { tp: 2, ep: 2, conc-start: 16, conc-end: 128 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 1024 }
-
-# Original config key: minimaxm3-fp4-b300-vllm
-minimaxm3-fp4-b300-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai:nightly-93d8f834dd8acf33eb0e2a75b2711b628cb6e226
-  model: nvidia/MiniMax-M3-NVFP4
-  model-prefix: minimaxm3
-  runner: b300
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 2 }
-      - { tp: 4, conc-start: 1, conc-end: 64 }
-      - { tp: 2, conc-start: 4, conc-end: 256 }
-      - { tp: 4, ep: 4, conc-start: 64, conc-end: 512 }
-      - { tp: 2, ep: 2, dp-attn: true, conc-start: 512, conc-end: 512 }
-      - { tp: 2, ep: 2, dp-attn: true, conc-start: 4096, conc-end: 4096 }
-
-# Original config key: minimaxm3-fp4-b300-vllm-mtp
-minimaxm3-fp4-b300-vllm-mtp-deprecated-1k1k:
-  image: vllm/vllm-openai:nightly-93d8f834dd8acf33eb0e2a75b2711b628cb6e226
-  model: nvidia/MiniMax-M3-NVFP4
-  model-prefix: minimaxm3
-  runner: b300
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 2, conc-list: [1, 2, 4, 8, 16, 32, 64, 512], spec-decoding: mtp }
-      - { tp: 2, ep: 2, conc-list: [1, 2, 4, 8, 16, 128, 512], spec-decoding: mtp }
-      - { tp: 2, ep: 2, dp-attn: true, conc-list: [256, 512], spec-decoding: mtp }
-      - { tp: 4, conc-list: [1, 2, 4, 8, 16, 32], spec-decoding: mtp }
-      - { tp: 4, ep: 4, conc-list: [1, 2, 32], spec-decoding: mtp }
-      - { tp: 8, conc-list: [1, 2, 4, 8], spec-decoding: mtp }
-
-# Original config key: minimaxm3-fp8-b200-vllm
-minimaxm3-fp8-b200-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai:minimax-m3-0618-x86_64-cu130
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: b200-dgxc
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 64 }
-      - { tp: 8, ep: 8, conc-start: 1, conc-end: 512 }
-      - { tp: 4, conc-start: 1, conc-end: 64 }
-      - { tp: 4, ep: 4, conc-start: 64, conc-end: 512 }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 128, conc-end: 512 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 1024 }
-
-# Original config key: minimaxm3-fp4-b200-vllm
-minimaxm3-fp4-b200-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai:nightly-93d8f834dd8acf33eb0e2a75b2711b628cb6e226
-  model: nvidia/MiniMax-M3-NVFP4
-  model-prefix: minimaxm3
-  runner: b200-dgxc
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 8 }
-      - { tp: 2, conc-start: 4, conc-end: 8 }
-      - { tp: 2, conc-start: 256, conc-end: 2048 }
-      - { tp: 4, conc-start: 32, conc-end: 256 }
-
-# Original config key: minimaxm3-fp8-b200-vllm-mtp
-minimaxm3-fp8-b200-vllm-mtp-deprecated-1k1k:
-  image: vllm/vllm-openai:minimax-m3-0618-x86_64-cu130
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: b200-dgxc
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 8, ep: 8, conc-start: 1, conc-end: 256, spec-decoding: mtp }
-      - { tp: 4, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 4, ep: 4, conc-start: 64, conc-end: 256, spec-decoding: mtp }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 128, conc-end: 512, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 512, spec-decoding: mtp }
-
-# Original config key: minimaxm3-fp4-b200-vllm-mtp
-minimaxm3-fp4-b200-vllm-mtp-deprecated-1k1k:
-  image: vllm/vllm-openai:vllm-minimax-m3-perf-x86_64-13.0.1-8b00f41
-  model: nvidia/MiniMax-M3-NVFP4
-  model-prefix: minimaxm3
-  runner: b200-dgxc
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 2, conc-list: [1, 2, 4, 8, 16, 64, 128, 256, 512], spec-decoding: mtp }
-      - { tp: 2, ep: 2, conc-list: [8, 256, 512], spec-decoding: mtp }
-      - { tp: 4, conc-list: [1, 2, 4, 8, 16, 32, 64, 256], spec-decoding: mtp }
-      - { tp: 4, ep: 4, conc-list: [1, 2, 4, 8, 256], spec-decoding: mtp }
-      - { tp: 8, conc-list: [1, 2, 4, 8, 16], spec-decoding: mtp }
-
-# Original config key: minimaxm3-fp8-b300-vllm-mtp
-minimaxm3-fp8-b300-vllm-mtp-deprecated-1k1k:
-  image: vllm/vllm-openai:minimax-m3-0618-x86_64-cu130
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: b300
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 8, ep: 8, conc-start: 1, conc-end: 256, spec-decoding: mtp }
-      - { tp: 4, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 4, ep: 4, conc-start: 64, conc-end: 256, spec-decoding: mtp }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 128, conc-end: 512, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 512, spec-decoding: mtp }
-
-# Original config key: minimaxm3-fp8-h200-vllm-mtp
-minimaxm3-fp8-h200-vllm-mtp-deprecated-1k1k:
-  image: vllm/vllm-openai:minimax-m3
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: h200
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 4, ep: 4, conc-start: 128, conc-end: 256, spec-decoding: mtp }
-      - { tp: 8, conc-start: 1, conc-end: 128, spec-decoding: mtp }
-      - { tp: 8, ep: 8, conc-start: 256, conc-end: 256, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 512, spec-decoding: mtp }
-
-# Original config key: minimaxm3-fp8-h100-vllm-mtp
-minimaxm3-fp8-h100-vllm-mtp-deprecated-1k1k:
-  image: vllm/vllm-openai:minimax-m3
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: h100
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 128, spec-decoding: mtp }
-      - { tp: 8, ep: 8, conc-start: 256, conc-end: 256, spec-decoding: mtp }
-
-# Original config key: kimik2.5-fp4-gb300-dynamo-vllm
-kimik2.5-fp4-gb300-dynamo-vllm-deprecated-1k1k:
-  image: vllm/vllm-openai:v0.21.0
-  model: nvidia/Kimi-K2.5-NVFP4
-  model-prefix: kimik2.5
-  runner: gb300-nv
-  precision: fp4
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.3.0.dev20260601" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - conc-list: [2048, 4096]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.5/vllm/gb300-fp4/1k1k/disagg-1p1d-dep4-dep16-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 1
-          ep: 16
-          dp-attn: true
-      - conc-list: [1024]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.5/vllm/gb300-fp4/1k1k/disagg-1p1d-dep4-dep24-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 1
-          ep: 24
-          dp-attn: true
-      - conc-list: [6144]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.5/vllm/gb300-fp4/1k1k/disagg-1p2d-dep4-dep4-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 1
-          ep: 4
-          dp-attn: true
-      - conc-list: [8, 16, 32, 64, 128]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.5/vllm/gb300-fp4/1k1k/disagg-1p7d-tep4-tp4-stp.yaml"
-        decode:
-          num-worker: 7
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [8192]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.5/vllm/gb300-fp4/1k1k/disagg-2p3d-dep4-dep8-stp.yaml"
-        decode:
-          num-worker: 3
-          tp: 1
-          ep: 8
-          dp-attn: true
-
-# Original config key: glm5-fp4-gb300-dynamo-trt-mtp
-glm5-fp4-gb300-dynamo-trt-mtp-deprecated-1k1k:
-  image: nvcr.io/nvidia/ai-dynamo/tensorrtllm-runtime:1.3.0-dev.1-cuda13
-  model: nvidia/GLM-5-NVFP4
-  model-prefix: glm5
-  runner: gb300-nv
-  precision: fp4
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "v1.3.0-dev.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # MTP configurations
-      - spec-decoding: "mtp"
-        conc-list: [ 8 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx1dep2_gen4tep8_batch1_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx1dep2_gen4tep8_batch1_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [ 12 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx1dep2_gen4tep8_batch2_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx1dep2_gen4tep8_batch2_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [ 24 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx1dep2_gen4tep8_batch4_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx1dep2_gen4tep8_batch4_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [ 84 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx1dep2_gen4tep8_batch16_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx1dep2_gen4tep8_batch16_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [ 180 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx1dep2_gen1dep32_batch4_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx1dep2_gen1dep32_batch4_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 333 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx1dep2_gen1dep32_batch8_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx1dep2_gen1dep32_batch8_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 666 ]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx2dep2_gen1dep16_batch32_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx2dep2_gen1dep16_batch32_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 1229 ]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx2dep2_gen1dep16_batch64_eplb0_mtp2.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx2dep2_gen1dep16_batch64_eplb0_mtp2.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 2253 ]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx3dep2_gen1dep16_batch128_eplb0_mtp1.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx3dep2_gen1dep16_batch128_eplb0_mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 4301 ]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx3dep2_gen1dep8_batch512_eplb0_mtp1.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx3dep2_gen1dep8_batch512_eplb0_mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 4301 ]
-        prefill:
-          num-worker: 4
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx4dep2_gen1dep16_batch256_eplb256_mtp1.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL1K_OSL1K/MTP/ctx4dep2_gen1dep16_batch256_eplb256_mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-qwen3.5-fp8-gb300-dynamo-sglang:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260709-074bb928
-  model: Qwen/Qwen3.5-397B-A17B-FP8
-  model-prefix: qwen3.5
-  runner: gb300
-  precision: fp8
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "46520ca59afe992fb5ef61b3197b2316f8df9b2b" }
-  kv-p2p-transfer: mooncake
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      # 1P1D STP: TP4 prefill + TP4 decode (pure tensor parallel). 2 nodes (1+1).
-      - spec-decoding: "none"
-        conc-list: [1, 2, 4, 8, 16, 32, 64]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/qwen3.5/sglang/gb300-fp8/1k1k/disagg-1p1d-tp4-tp4-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1P1D wide-EP: prefill DEP4 + decode DEP16. 5 nodes (1+4).
-      - spec-decoding: "none"
-        conc-list: [512, 1024, 2048]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/qwen3.5/sglang/gb300-fp8/1k1k/disagg-1p1d-dep4-dep16-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # 2P1D wide-EP: 2 prefill DEP4 + decode DEP16. 6 nodes (2+4).
-      - spec-decoding: "none"
-        conc-list: [4096]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/qwen3.5/sglang/gb300-fp8/1k1k/disagg-2p1d-dep4-dep16-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-# Archive source: nvidia-dsv4-8k1k-master.yaml
-# Deprecated DeepSeek-V4-Pro 1.6T Single-turn 8k1k entries archived from
-# nvidia-master.yaml. Deprecated after Tuesday, September 8, 2026 (see the
-# Deprecation Notice in MODELS.md). Agentic coding is unaffected and stays
-# active for dsv4, including its MTP and DSpark arms; the model is not retired.
-# Removed from the active master config so sweep generation no longer selects them.
-
-dsv4-fp4-b200-sglang:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260628-da802ddc
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: cluster:b200-nscale
-  precision: fp4
-  framework: sglang
-  multinode: false
-  # Two recipes from https://docs.sglang.io/cookbook/autoregressive/DeepSeek/DeepSeek-V4
-  # are selected inside benchmarks/single_node/dsv4_fp4_b200.sh by DP_ATTENTION:
-  #   low-latency  (DP_ATTENTION=false): TP-only, flashinfer_mxfp4
-  #   DP-attention  (DP_ATTENTION=true):  DP-attn + DeepEP + mega_moe opts
-  # The DP-attention recipe covers both "balanced" (conc 64-128) and
-  # "max-throughput" (conc 256+) CONC ranges with identical flags;
-  # only --max-running-requests scales with CONC.
-  # ep is implicit in sglang: --moe-a2a-backend deepep forces ep_size=tp_size,
-  # while low-latency leaves ep_size at the default of 1.
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # low-latency (DP_ATTENTION=false)
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 32 }
-      # DP-attention (DP_ATTENTION=true) — balanced CONC range
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 64, conc-end: 128 }
-      # DP-attention (DP_ATTENTION=true) — max-throughput CONC range
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 1024 }
-
-dsv4-fp4-b200-vllm:
-  image: vllm/vllm-openai:nightly-7c5dc571cbd1064ecc8a9b1045637ff647aa22cb
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: cluster:b200-nscale
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 256 }
-      - { tp: 8, ep: 8, conc-start: 1, conc-end: 256 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 64, conc-end: 1024 }
-
-dsv4-fp4-b200-trt:
-  image: ghcr.io#semianalysisai/trtllm-deepseek-v4:feat-deepseek_v4-c185066
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: cluster:b200-nscale
-  precision: fp4
-  framework: trt
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 32 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 32, conc-end: 512 }
-
-dsv4-fp4-b200-trt-mtp:
-  image: ghcr.io#semianalysisai/trtllm-deepseek-v4:feat-deepseek_v4-c185066
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: cluster:b200-nscale
-  precision: fp4
-  framework: trt
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 32, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 32, conc-end: 256, spec-decoding: mtp }
-
-# MTP variant of dsv4-fp4-b200-vllm. Mirrors the base search space and adds
-# --speculative-config '{"method":"mtp","num_speculative_tokens":2}'.
-dsv4-fp4-b200-vllm-mtp:
-  image: vllm/vllm-openai:v0.25.0
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: cluster:b200-nscale
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 8, ep: 8, conc-start: 128, conc-end: 256, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 64, conc-end: 256, spec-decoding: mtp }
-
-# DeepSeek-V4-Pro on B300 with sglang (non-MTP).
-# Uses nightly image with megamoe backend for high-concurrency profiles.
-dsv4-fp4-b300-sglang:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260624-b2c8f7a2
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: b300
-  precision: fp4
-  framework: sglang
-  multinode: false
-  # Recipes are selected inside benchmarks/single_node/dsv4_fp4_b300_sglang.sh
-  # by CONC:
-  #   CONC 1|32:         TP-only, flashinfer_mxfp4
-  #   CONC 512:          DP-attn, flashinfer_mxfp4
-  #   CONC 2048-8192:    DP-attn, megamoe
-  # ep is implicit in sglang: --moe-a2a-backend megamoe forces ep_size=tp_size,
-  # while low-latency leaves ep_size at the default of 1.
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 1 }
-      - { tp: 4, ep: 1, conc-start: 32, conc-end: 32 }
-      - { tp: 4, ep: 1, dp-attn: true, conc-start: 512, conc-end: 512 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 2048, conc-end: 2048 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 4096, conc-end: 4096 }
-
-dsv4-fp4-b300-sglang-mtp:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260610-f332e526
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: b300
-  precision: fp4
-  framework: sglang
-  multinode: false
-  # Three CONC bands:
-  #   A: TP=8 ep=1            -- conc 1-8    EAGLE (3,1,4) TP-only fallback
-  #   B: TP=4 ep=1            -- conc 4-32   EAGLE (3,1,4) TP-only mid batch
-  #   C: TP=4 ep=1 dp-attn    -- conc 16-256 EAGLE (1,1,2) DP-attn flashinfer
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 8, spec-decoding: mtp }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 32, spec-decoding: mtp }
-
-# DeepSeek-V4-Pro H200 recipe from https://vllm.ai/blog/deepseek-v4
-# Uses the cu129 image. H200 has no FP4 path, so the FP4 indexer cache
-# flag is omitted. Max-model-len is pinned at 800k per the recipe.
-dsv4-fp8-h200-vllm:
-  image: vllm/vllm-openai:v0.25.1
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: h200
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, dp-attn: false, conc-start: 1, conc-end: 256 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 1, conc-end: 256 }
-
-# MTP variant of dsv4-fp8-h200-vllm. Mirrors the non-MTP image and adds
-# --speculative-config '{"method":"mtp","num_speculative_tokens":2}'.
-dsv4-fp8-h200-vllm-mtp:
-  image: vllm/vllm-openai:v0.25.1
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: h200
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, dp-attn: false, conc-start: 1, conc-end: 256, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 1, conc-end: 256, spec-decoding: mtp }
-
-# DeepSeek-V4-Pro H200 single-node with SGLang (Marlin FP8, TP-only).
-# Pinned to the h200-dgxc-slurm runner pool because the deepseek-v4-hopper
-# image needs the /ix mount layout that only launch_h200-dgxc-slurm.sh sets up.
-
-# MTP variant of dsv4-fp8-h200-sglang. Mirrors the non-MTP recipe (same image,
-# runner pool, search space) and adds EAGLE speculative decoding via
-# --speculative-algorithm EAGLE with the (3,1,4) chain matching dsv4-fp4-b300-sglang-mtp.
-dsv4-fp8-h200-sglang:
-  image: lmsysorg/sglang:deepseek-v4-hopper@sha256:1bf5d508ab110cc0fe1659a5f21d1be02a7f0d7ca8f58cea7e7f4e11f6ae208f
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: h200-dgxc
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 1 }
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64 }
-
-# MTP variant of dsv4-fp8-h200-sglang. Mirrors the non-MTP recipe (same image,
-# runner pool, search space) and adds EAGLE speculative decoding via
-# --speculative-algorithm EAGLE with the (3,1,4) chain matching dsv4-fp4-b300-sglang-mtp.
-dsv4-fp8-h200-sglang-mtp:
-  image: lmsysorg/sglang:deepseek-v4-hopper@sha256:1bf5d508ab110cc0fe1659a5f21d1be02a7f0d7ca8f58cea7e7f4e11f6ae208f
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: h200-dgxc
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 1, spec-decoding: mtp }
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-
-  # DeepSeek-V4-Pro B300 single-node aggregate recipe from the submitted B300
-  # pareto sweep. The single-node schema has no explicit data-parallel-size
-  # field, so dp-attn=true is used as the existing vLLM script switch for DP4
-  # layouts on 4 allocated GPUs.
-
-dsv4-fp4-b300-vllm:
-  image: vllm/vllm-openai:v0.25.0
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: cluster:b300-nv
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 64 }
-      - { tp: 8, conc-start: 1, conc-end: 4 }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 256, conc-end: 512 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 2048, conc-end: 2048 }
-
-dsv4-fp4-b300-trt:
-  image: ghcr.io#semianalysisai/trtllm-deepseek-v4:feat-deepseek_v4-c185066
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: b300
-  precision: fp4
-  framework: trt
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 32 }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 64, conc-end: 64 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 1024 }
-
-dsv4-fp4-b300-trt-mtp:
-  image: ghcr.io#semianalysisai/trtllm-deepseek-v4:feat-deepseek_v4-c185066
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: b300
-  precision: fp4
-  framework: trt
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 32, spec-decoding: mtp }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 64, conc-end: 64, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 1024, spec-decoding: mtp }
-
-dsv4-fp4-b300-vllm-mtp:
-  image: vllm/vllm-openai:v0.25.0
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: cluster:b300-nv
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 8, conc-start: 1, conc-end: 8, spec-decoding: mtp }
-      - { tp: 4, ep: 4, conc-start: 64, conc-end: 256, spec-decoding: mtp }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 256, conc-end: 512, spec-decoding: mtp }
-
-dsv4-fp4-b200-dynamo-vllm:
-  image: vllm/vllm-openai:v0.23.0
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: cluster:b200-nscale
-  precision: fp4
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.2.0.dev20260426" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # B200 adaptation of the DSV4 GB200 vLLM disagg recipes. Each worker
-      # maps to one full 8-GPU B200 node.
-      - conc-list: [1]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/b200-fp4/8k1k/disagg-1p1d-p-tp1-dp8-d-tp8-c1-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [32, 128]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/b200-fp4/8k1k/disagg-1p1d-p-tp1-dp8-d-tp8-c32x128-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [64]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/b200-fp4/8k1k/disagg-1p1d-p-tp1-dp8-d-tp8-c64-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [256]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/b200-fp4/8k1k/disagg-1p1d-p-tp1-dp8-d-tp1-dp8-c256-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [512]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/b200-fp4/8k1k/disagg-1p1d-p-tp1-dp8-d-tp1-dp8-c512-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-dsv4-fp4-gb200-dynamo-vllm:
-  image: vllm/vllm-openai:v0.25.1
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: gb200
-  precision: fp4
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.2.0.dev20260426" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # Validated 8k/1k points mirrored from NVIDIA/srt-slurm
-      # aflowers/vllm-gb200-v0.20.0 history. conc-list values match each
-      # recipe's benchmark.concurrencies.
-
-      # Low latency: 1 prefill (DEP=8) + 1 decode (TP=8). 5 nodes total with
-      # a dedicated NATS/etcd infra node.
-      - conc-list: [1]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/gb200-fp4/8k1k/disagg-1p1d-p-tp1-dp8-d-tp8-c1-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-
-      # Low-middle curve: 1 prefill (DEP=8) + 4 decode (TP=8). 11 nodes total
-      # with a dedicated NATS/etcd infra node.
-      - conc-list: [256, 512]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/gb200-fp4/8k1k/disagg-1p4d-p-tp1-dp8-d-tp8-c256x512-stp.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 1
-          dp-attn: false
-
-      # MegaMOE mid curve: 1 prefill (DEP=8) + 1 decode (DEP=8). 5 nodes
-      # total with a dedicated NATS/etcd infra node.
-      - conc-list: [256, 512, 1024]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/gb200-fp4/8k1k/disagg-1p1d-p-tp1-dp8-d-tp1-dp8-c256x512x1024-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-      # MegaMOE high throughput: 2 prefill (DEP=8 each) + 1 decode (DEP=8).
-      # 7 nodes total with a dedicated NATS/etcd infra node.
-      - conc-list: [4096]
-        prefill:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/gb200-fp4/8k1k/disagg-high-tpt-megamoe-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-      # MegaMOE max throughput: 3 prefill (DEP=8 each) + 1 decode (DEP=8).
-      # 9 nodes total with a dedicated NATS/etcd infra node.
-      - conc-list: [4096]
-        prefill:
-          num-worker: 3
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/gb200-fp4/8k1k/disagg-3p1d-p-tp1-dp8-d-tp1-dp8-c4096-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-# TODO: change image to official llmd image.
-# Build source: benchmarks/llm-d/Dockerfile.
-dsv4-fp4-gb200-llmd-vllm:
-  image: quay.io/rh-ee-imarkov/llm-d-nokube-vllm:vllm0.26@sha256:a9095d4c835935c4070be2040de0a5ef3b44098f603092ab66d743b0e731b7b4
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: gb200
-  precision: fp4
-  framework: llmd-vllm
-  router: { name: llm-d-router, version: "0.9.0" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # Low latency: 1 prefill DEP8 + 1 decode TP8.
-      - spec-decoding: "none"
-        conc-list: [1]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "PREFILL_NODES=2"
-          - "GPUS_PER_NODE=4"
-          - "CONFIG_FILE=dsv4-fp4-gb200-low-latency.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "GPUS_PER_NODE=4"
-
-      # Mid curve: 1 prefill DEP8 + 1 decode DEP8.
-      - spec-decoding: "none"
-        conc-list: [256, 512, 1024]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "PREFILL_NODES=2"
-          - "GPUS_PER_NODE=4"
-          - "CONFIG_FILE=dsv4-fp4-gb200-mid-curve-megamoe.yaml"
-        decode:
-          num-worker: 1
-          tp: 1
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "GPUS_PER_NODE=4"
-
-      # Max throughput: 3 prefill DEP8 + 1 decode DEP8.
-      - spec-decoding: "none"
-        conc-list: [4096]
-        prefill:
-          num-worker: 3
-          tp: 1
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "PREFILL_NODES=6"
-          - "PREFILL_WORKERS=3"
-          - "GPUS_PER_NODE=4"
-          - "CONFIG_FILE=dsv4-fp4-gb200-mid-curve-megamoe.yaml"
-        decode:
-          num-worker: 1
-          tp: 1
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "GPUS_PER_NODE=4"
-
-# MTP2 variant of dsv4-fp4-gb200-dynamo-vllm. Uses the vLLM 0.20.1 image
-# and hand-picked 8k/1k Pareto points mirrored from NVIDIA/srt-slurm.
-dsv4-fp4-gb200-dynamo-vllm-mtp2:
-  image: vllm/vllm-openai:v0.20.1-ubuntu2404
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: gb200
-  precision: fp4
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.2.0.dev20260426" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # Aggregate low latency: TP=8, max-num-seqs=4.
-      - conc-list: [1]
-        spec-decoding: mtp
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "SYNTHETIC_ACCEPTANCE=true"
-          - "SYNTHETIC_ACCEPTANCE_LENGTH=2.27"
-          - "CONFIG_FILE=recipes/dsv4/vllm/gb200-fp4/8k1k/agg-tp8-c1-mtp.yaml"
-        decode:
-          num-worker: 0
-          tp: 8
-          ep: 1
-          dp-attn: false
-
-      # Low-latency bridge: 1 prefill (DEP=8) + 4 decode (TP=8), no offload.
-      - conc-list: [16, 32, 64]
-        spec-decoding: mtp
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "SYNTHETIC_ACCEPTANCE=true"
-          - "SYNTHETIC_ACCEPTANCE_LENGTH=2.27"
-          - "CONFIG_FILE=recipes/dsv4/vllm/gb200-fp4/8k1k/disagg-1p4d-p-tp1-dp8-d-tp8-c16x32x64-mtp.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 1
-          dp-attn: false
-
-      # MegaMOE mid curve: 1 prefill (DEP=8) + 1 decode (DEP=8).
-      # 5 nodes total with a dedicated NATS/etcd infra node.
-      - conc-list: [128, 256, 512, 1024]
-        spec-decoding: mtp
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "SYNTHETIC_ACCEPTANCE=true"
-          - "SYNTHETIC_ACCEPTANCE_LENGTH=2.27"
-          - "CONFIG_FILE=recipes/dsv4/vllm/gb200-fp4/8k1k/disagg-1p1d-p-tp1-dp8-d-tp1-dp8-c128x256x512x1024-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-      # MegaMOE high throughput: 2 prefill (DEP=8 each) + 1 decode (DEP=8).
-      # 7 nodes total with a dedicated NATS/etcd infra node.
-      - conc-list: [1024]
-        spec-decoding: mtp
-        prefill:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "SYNTHETIC_ACCEPTANCE=true"
-          - "SYNTHETIC_ACCEPTANCE_LENGTH=2.27"
-          - "CONFIG_FILE=recipes/dsv4/vllm/gb200-fp4/8k1k/disagg-high-tpt-megamoe-mtp2.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-# Baseline (no synthetic acceptance) variant of dsv4-fp4-gb200-dynamo-vllm-mtp2
-# for before/after Pareto comparison. Identical topology and recipes, only the
-# SYNTHETIC_ACCEPTANCE envs are removed.
-dsv4-fp4-gb200-dynamo-vllm-mtp2-nosynthetic:
-  image: vllm/vllm-openai:v0.20.1-ubuntu2404
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: gb200
-  precision: fp4
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.2.0.dev20260426" }
-  multinode: true
-  disagg: true
-  kv-p2p-transfer: nixl
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # Aggregate low latency: TP=8, max-num-seqs=4.
-      - conc-list: [1]
-        spec-decoding: mtp
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/gb200-fp4/8k1k/agg-tp8-c1-mtp.yaml"
-        decode:
-          num-worker: 0
-          tp: 8
-          ep: 1
-          dp-attn: false
-
-      # Low-latency bridge: 1 prefill (DEP=8) + 4 decode (TP=8), no offload.
-      - conc-list: [16, 32, 64]
-        spec-decoding: mtp
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/gb200-fp4/8k1k/disagg-1p4d-p-tp1-dp8-d-tp8-c16x32x64-mtp.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 1
-          dp-attn: false
-
-      # MegaMOE mid curve: 1 prefill (DEP=8) + 1 decode (DEP=8).
-      # 5 nodes total with a dedicated NATS/etcd infra node.
-      - conc-list: [128, 256, 512, 1024]
-        spec-decoding: mtp
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/gb200-fp4/8k1k/disagg-1p1d-p-tp1-dp8-d-tp1-dp8-c128x256x512x1024-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-      # MegaMOE high throughput: 2 prefill (DEP=8 each) + 1 decode (DEP=8).
-      # 7 nodes total with a dedicated NATS/etcd infra node.
-      - conc-list: [1024]
-        spec-decoding: mtp
-        prefill:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/gb200-fp4/8k1k/disagg-high-tpt-megamoe-mtp2.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-# DSV4 B200 disaggregated Dynamo SGLang STP configuration.
-dsv4-fp4-b200-dynamo-sglang:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260821-f825d729
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: cluster:b200-nscale
-  precision: fp4
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "86f84b9486905c2bfaa4b756c9fa39f04ecf5ebd" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 1p1d-tp8-tp8 (pure TP). 2 nodes (1+1).
-      - conc-list: [1]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b200-fp4/8k1k/disagg-1p1d-tp8-tp8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # 1p1d-dep8-tp8: DEP8 prefill + TP8 decode. 2 nodes (1+1).
-      - conc-list: [32, 64]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b200-fp4/8k1k/disagg-1p1d-dep8-tp8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # 1p1d-dep8-dep8: DEP8 prefill + DEP8 decode. 2 nodes (1+1).
-      - conc-list: [256, 512]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b200-fp4/8k1k/disagg-1p1d-dep8-dep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # 1p4d-dep8-tp8: DEP8 prefill + 4x TP8 decode. 5 nodes (1+4).
-      - conc-list: [64]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b200-fp4/8k1k/disagg-1p4d-dep8-tp8-stp.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # 1p2d-dep8-dep8: DEP8 prefill + 2x DEP8 decode. 3 nodes (1+2).
-      - conc-list: [256]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b200-fp4/8k1k/disagg-1p2d-dep8-dep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # 1p4d-dep8-dep8: DEP8 prefill + 4x DEP8 decode. 5 nodes (1+4).
-      - conc-list: [256]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b200-fp4/8k1k/disagg-1p4d-dep8-dep8-stp.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # 3p2d-dep8-dep8: 3x DEP8 prefill + 2x DEP8 decode. 5 nodes (3+2).
-      - conc-list: [2048]
-        prefill:
-          num-worker: 3
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b200-fp4/8k1k/disagg-3p2d-dep8-dep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # 4p2d-dep8-dep8: 4x DEP8 prefill + 2x DEP8 decode. 6 nodes (4+2).
-      - conc-list: [4096]
-        prefill:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b200-fp4/8k1k/disagg-4p2d-dep8-dep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # 5p2d-dep8-dep8: 5x DEP8 prefill + 2x DEP8 decode. 7 nodes (5+2).
-      - conc-list: [6144]
-        prefill:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b200-fp4/8k1k/disagg-5p2d-dep8-dep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # 6p2d-dep8-dep8: 6x DEP8 prefill + 2x DEP8 decode. 8 nodes (6+2).
-      - conc-list: [8192]
-        prefill:
-          num-worker: 6
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b200-fp4/8k1k/disagg-6p2d-dep8-dep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-dsv4-fp4-gb200-dynamo-sglang:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260528-0abe6a85
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: gb200
-  precision: fp4
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "92f5b3b8d7dd5ab9179d4b1034bd2c1c0803693e" }
-  kv-p2p-transfer: mooncake
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # Low latency: 1p1d-tp8-tp8. 4 nodes.
-      - conc-list: [1]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb200-fp4/8k1k/disagg-1p1d-tp8-tp8-4-c1-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # 1p4d-dep8-tp8. 10 nodes.
-      - conc-list: [64]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb200-fp4/8k1k/disagg-1p4d-dep8-tp8-10-c64-stp.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # WideEP TP=16 decode: 1p2d-dep8-dep16. 10 nodes.
-      - conc-list: [256]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb200-fp4/8k1k/disagg-1p2d-dep8-dep16-10-c256-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # WideEP TP=16 decode: 1p1d-dep8-dep16. 6 nodes.
-      - conc-list: [512]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb200-fp4/8k1k/disagg-1p1d-dep8-dep16-6-c512-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # WideEP TP=16 decode: 2p1d-dep8-dep16. 8 nodes.
-      - conc-list: [1536]
-        prefill:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb200-fp4/8k1k/disagg-2p1d-dep8-dep16-8-c1536-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # WideEP TP=16 decode: 4p1d-dep8-dep16. 12 nodes.
-      - conc-list: [4096]
-        prefill:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb200-fp4/8k1k/disagg-4p1d-dep8-dep16-12-c4096-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # WideEP TP=16 decode: 5p1d-dep8-dep16. 14 nodes.
-      - conc-list: [8192]
-        prefill:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb200-fp4/8k1k/disagg-5p1d-dep8-dep16-14-c8192-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # WideEP TP=12 decode: 6p1d-dep8-dep12. 15 nodes.
-      - conc-list: [8192]
-        prefill:
-          num-worker: 6
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb200-fp4/8k1k/disagg-6p1d-dep8-dep12-15-c8192-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 12
-          ep: 12
-          dp-attn: true
-
-# MTP variant of dsv4-fp4-gb200-dynamo-sglang.
-dsv4-fp4-gb200-dynamo-sglang-mtp:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260528-0abe6a85
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: gb200
-  precision: fp4
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "92f5b3b8d7dd5ab9179d4b1034bd2c1c0803693e" }
-  kv-p2p-transfer: mooncake
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # Low-latency baseline: 1p1d-tp8-tp8. 4 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [1]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb200-fp4/8k1k/disagg-1p1d-p-tp8-d-tp8-b8-c1-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # Low-latency 1p6d-dep8-tp8: 1P (DEP=8) + 6 TP=8 decode workers. 14 nodes.
-      # Recipe runs concurrencies=32x64x128; matrix tracks the max.
-      - spec-decoding: "mtp"
-        conc-list: [128]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb200-fp4/8k1k/disagg-1p6d-p-tp8-ep8-dp8-d-tp8-b128-c32x64x128-mtp.yaml"
-        decode:
-          num-worker: 6
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # Mid curve 1p1d-dep8-dep16. 6 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [1024]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb200-fp4/8k1k/disagg-1p1d-p-tp8-ep8-dp8-d-tp16-ep16-dp16-b1024-c1024-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # Mid curve 2p1d-dep8-dep16. 8 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [2048]
-        prefill:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb200-fp4/8k1k/disagg-2p1d-p-tp8-ep8-dp8-d-tp16-ep16-dp16-b2048-c2048-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # Mid curve 3p1d-dep8-dep16. 10 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [3072]
-        prefill:
-          num-worker: 3
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb200-fp4/8k1k/disagg-3p1d-p-tp8-ep8-dp8-d-tp16-ep16-dp16-b4096-c3072-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # Mid curve 4p1d-dep8-dep16. 12 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [6144]
-        prefill:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb200-fp4/8k1k/disagg-4p1d-p-tp8-ep8-dp8-d-tp16-ep16-dp16-b6144-c6144-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # Mid curve 5p1d-dep8-dep16. 14 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [8192]
-        prefill:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb200-fp4/8k1k/disagg-5p1d-p-tp8-ep8-dp8-d-tp16-ep16-dp16-b16384-c8192-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # Mid curve 6p1d-dep8-dep16. 16 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [16384]
-        prefill:
-          num-worker: 6
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb200-fp4/8k1k/disagg-6p1d-p-tp8-ep8-dp8-d-tp16-ep16-dp16-b21504-c16384-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-    # ---------- 1k1k high-throughput (wide-EP decode, MTP) ----------
-
-dsv4-fp4-b300-dynamo-vllm:
-  image: vllm/vllm-openai:v0.23.0
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: b300
-  precision: fp4
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.2.0.dev20260426" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # B300 adaptation of the DSV4 B200/GB200 vLLM disagg recipes. Each
-      # prefill/decode worker maps to one full 8-GPU B300 node.
-      - conc-list: [1, 32, 64, 128]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/b300-fp4/8k1k/disagg-1p1d-p-tp1-dp8-d-tp8-c1x32x64x128-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [256, 1024]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/b300-fp4/8k1k/disagg-1p1d-p-tp1-dp8-d-tp1-dp8-c256x1024-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [4096]
-        prefill:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/b300-fp4/8k1k/disagg-high-tpt-megamoe-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-# DSV4 B300 disaggregated dynamo-sglang STP configuration.
-# Each search-space entry maps one-to-one to a checked-in srt-slurm recipe under
-# recipes/sglang/deepseek-v4/8k1k/. Runs on the b300 pool via
-# runners/launch_b300-nv.sh. Prefill workers are TP4/DEP4 (two workers per
-# 8-GPU B300 node); the decode worker is a single DEP8 node.
-dsv4-fp4-b300-dynamo-sglang:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260721-8905cbd4
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: b300
-  precision: fp4
-  framework: dynamo-sglang
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 1p1d-tp4-tp4 (pure TP, no dp-attn). 2 nodes (1P + 1D).
-      - conc-list: [1, 8, 16, 32]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b300-fp4/8k1k/disagg-1p1d-tp4-tp4-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p1d-dep4-dep8: 1x DEP4 prefill + DEP8 decode. 2 nodes (1P + 1D).
-      - conc-list: [256]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b300-fp4/8k1k/disagg-1p1d-dep4-dep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # 2p1d-dep4-dep8: 2x DEP4 prefill + DEP8 decode. 2 nodes (1P + 1D).
-      - conc-list: [768]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b300-fp4/8k1k/disagg-2p1d-dep4-dep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # 4p1d-dep4-dep8: 4x DEP4 prefill + DEP8 decode. 3 nodes (2P + 1D).
-      - conc-list: [2048]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b300-fp4/8k1k/disagg-4p1d-dep4-dep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # 6p1d-dep4-dep8: 6x DEP4 prefill + DEP8 decode. 4 nodes (3P + 1D).
-      - conc-list: [3072]
-        prefill:
-          num-worker: 6
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b300-fp4/8k1k/disagg-6p1d-dep4-dep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-dsv4-fp4-gb300-dynamo-vllm:
-  image: vllm/vllm-openai:dsv4-megamoe-mxfp4-arm64-cu130-4ba0a72
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: gb300-nv
-  precision: fp4
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.2.0.dev20260426" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [192]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/gb300-fp4/8k1k/disagg-1p6d-dep4-tp4-stp.yaml"
-        decode:
-          num-worker: 6
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [18]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/gb300-fp4/8k1k/disagg-1p9d-tep4-tp4-stp.yaml"
-        decode:
-          num-worker: 9
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [4096]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/gb300-fp4/8k1k/disagg-4p1d-dep4-dep8-24-c4096-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [4096]
-        prefill:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/gb300-fp4/8k1k/disagg-5p1d-dep4-dep8-28-c4096-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [4096]
-        prefill:
-          num-worker: 6
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/gb300-fp4/8k1k/disagg-6p1d-dep4-dep8-32-c4096-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [3072]
-        prefill:
-          num-worker: 7
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/vllm/gb300-fp4/8k1k/disagg-7p2d-dep4-dep16-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-dsv4-fp4-b200-dynamo-sglang-mtp:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260710-cfc66e05
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: cluster:b200-nscale
-  precision: fp4
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "92f5b3b8d7dd5ab9179d4b1034bd2c1c0803693e" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - spec-decoding: "mtp"
-        conc-list: [1]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b200-fp4/8k1k/disagg-1p1d-p-tp8-d-tp8-b8-c1-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [32, 64, 128]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b200-fp4/8k1k/disagg-1p6d-p-tp8-ep8-dp8-d-tp8-b128-c32x64x128-mtp.yaml"
-        decode:
-          num-worker: 6
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [256, 1024]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b200-fp4/8k1k/disagg-1p1d-p-tp8-ep8-dp8-d-tp8-ep8-dp8-b1536-c256x1024-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [256]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b200-fp4/8k1k/disagg-1p2d-p-tp8-ep8-dp8-d-tp8-ep8-dp8-b1024-c256-mtp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [6144]
-        prefill:
-          num-worker: 5
-          tp: 8
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b200-fp4/8k1k/disagg-5p3d-p-tp8-ep8-dp8-d-tp8-ep8-dp8-b2560-c6144-mtp.yaml"
-        decode:
-          num-worker: 3
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-dsv4-fp4-b200-dynamo-vllm-mtp:
-  image: vllm/vllm-openai:vllm-x86_64-cu13-0.25.1-7a33ba9
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: cluster:b200-nscale
-  precision: fp4
-  framework: dynamo-vllm
-  multinode: true
-  disagg: true
-  kv-p2p-transfer: nixl
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # B200 MTP disagg recipes. Each worker maps to one full 8-GPU B200 node
-      # (prefill dep8; decode tp8, except the low-latency tp8 variant).
-      # SYNTHETIC_ACCEPTANCE is a throughput-only control (injected into the
-      # recipe by runners/inject_synthetic_acceptance.py); EVAL_ONLY runs keep
-      # the checked-in real-MTP recipe so verification is measured normally.
-      - spec-decoding: "mtp"
-        conc-list: [1, 8, 16]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "SYNTHETIC_ACCEPTANCE=true"
-          - "SYNTHETIC_ACCEPTANCE_LENGTH=2.49"
-          - "CONFIG_FILE=recipes/dsv4/vllm/b200-fp4/8k1k/disagg-1p1d-dep8-tp8-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [128, 256]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "SYNTHETIC_ACCEPTANCE=true"
-          - "SYNTHETIC_ACCEPTANCE_LENGTH=2.49"
-          - "CONFIG_FILE=recipes/dsv4/vllm/b200-fp4/8k1k/disagg-1p2d-dep8-dep8-mtp.yaml"
-        decode:
-          num-worker: 2
-          tp: 1
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [512]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "SYNTHETIC_ACCEPTANCE=true"
-          - "SYNTHETIC_ACCEPTANCE_LENGTH=2.49"
-          - "CONFIG_FILE=recipes/dsv4/vllm/b200-fp4/8k1k/disagg-1p1d-dep8-dep8-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 1
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [1024]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "SYNTHETIC_ACCEPTANCE=true"
-          - "SYNTHETIC_ACCEPTANCE_LENGTH=2.49"
-          - "CONFIG_FILE=recipes/dsv4/vllm/b200-fp4/8k1k/disagg-2p1d-dep8-dep8-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 1
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [4096]
-        prefill:
-          num-worker: 3
-          tp: 1
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "SYNTHETIC_ACCEPTANCE=true"
-          - "SYNTHETIC_ACCEPTANCE_LENGTH=1.79"
-          - "CONFIG_FILE=recipes/dsv4/vllm/b200-fp4/8k1k/disagg-3p1d-dep8-dep8-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 1
-          ep: 8
-          dp-attn: true
-
-dsv4-fp4-gb300-dynamo-trt:
-  image: nvcr.io#nvidia/ai-dynamo/tensorrtllm-runtime:1.3.0-deepseek-v4-dev.1
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: gb300-nv
-  precision: fp4
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "v1.3.0-deepseek-v4-dev.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [4]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/STP/ctx1dep4_gen4tep8_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-1p4d-dep4-tep8-b1-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [5]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-1p5d-dep4-tep4-b1-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [15]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch2_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-1p5d-dep4-tep4-b2-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [25]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch4_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-1p5d-dep4-tep4-b4-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [55]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch8_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-1p5d-dep4-tep4-b8-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [154]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/STP/ctx1dep4_gen1dep32_batch4_eplb384_mtp0.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-1p1d-dep4-dep32-b4-eplb384-mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [308]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/STP/ctx2dep4_gen1dep32_batch8_eplb384_mtp0.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-2p1d-dep4-dep32-b8-eplb384-mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [615]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/STP/ctx4dep4_gen1dep32_batch16_eplb384_mtp0.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-4p1d-dep4-dep32-b16-eplb384-mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [1127]
-        prefill:
-          num-worker: 6
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/STP/ctx6dep4_gen1dep32_batch32_eplb384_mtp0.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-6p1d-dep4-dep32-b32-eplb384-mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [1229]
-        prefill:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/STP/ctx5dep4_gen1dep16_batch64_eplb384_mtp0.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-5p1d-dep4-dep16-b64-eplb384-mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [2253]
-        prefill:
-          num-worker: 6
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/STP/ctx6dep4_gen1dep8_batch256_eplb384_mtp0.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-6p1d-dep4-dep8-b256-eplb384-mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [2253]
-        prefill:
-          num-worker: 9
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/STP/ctx9dep4_gen1dep16_batch128_eplb384_mtp0.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-9p1d-dep4-dep16-b128-eplb384-mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [4301]
-        prefill:
-          num-worker: 10
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/STP/ctx10dep4_gen1dep8_batch512_eplb384_mtp0.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-10p1d-dep4-dep8-b512-eplb384-mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-dsv4-fp4-gb300-dynamo-trt-mtp:
-  image: nvcr.io#nvidia/ai-dynamo/tensorrtllm-runtime:1.3.0-deepseek-v4-dev.1
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: gb300-nv
-  precision: fp4
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "v1.3.0-deepseek-v4-dev.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - spec-decoding: "mtp"
-        conc-list: [8]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/MTP/ctx1dep4_gen4tep8_batch1_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-1p4d-dep4-tep8-b1-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [10]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/MTP/ctx1dep4_gen5tep4_batch1_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-1p5d-dep4-tep4-b1-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [15]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/MTP/ctx1dep4_gen5tep4_batch2_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-1p5d-dep4-tep4-b2-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [30]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/MTP/ctx1dep4_gen5tep4_batch4_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-1p5d-dep4-tep4-b4-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [84]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/MTP/ctx2dep4_gen1dep32_batch2_eplb384_mtp3.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-2p1d-dep4-dep32-b2-eplb384-mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [180]
-        prefill:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/MTP/ctx3dep4_gen1dep32_batch4_eplb384_mtp3.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-3p1d-dep4-dep32-b4-eplb384-mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [333]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/MTP/ctx4dep4_gen1dep32_batch8_eplb384_mtp3.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-4p1d-dep4-dep32-b8-eplb384-mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [615]
-        prefill:
-          num-worker: 8
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/MTP/ctx8dep4_gen1dep32_batch16_eplb384_mtp3.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-8p1d-dep4-dep32-b16-eplb384-mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [666]
-        prefill:
-          num-worker: 6
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/MTP/ctx6dep4_gen1dep16_batch32_eplb384_mtp3.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-6p1d-dep4-dep16-b32-eplb384-mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [1229]
-        prefill:
-          num-worker: 7
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/MTP/ctx7dep4_gen1dep8_batch128_eplb384_mtp3.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-7p1d-dep4-dep8-b128-eplb384-mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [1229]
-        prefill:
-          num-worker: 10
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/MTP/ctx10dep4_gen1dep16_batch64_eplb384_mtp3.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-10p1d-dep4-dep16-b64-eplb384-mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [2253]
-        prefill:
-          num-worker: 9
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/MTP/ctx9dep4_gen1dep8_batch256_eplb384_mtp1.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-9p1d-dep4-dep8-b256-eplb384-mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [4301]
-        prefill:
-          num-worker: 12
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/DeepSeek-V4-Pro/disagg/trtllm_dynamo/gb300_mxfp4/ISL8K_OSL1K/MTP/ctx12dep4_gen1dep8_batch512_eplb384_mtp1.yaml
-          - "CONFIG_FILE=recipes/dsv4/trtllm/gb300-fp4/8k1k/disagg-12p1d-dep4-dep8-b512-eplb384-mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-dsv4-fp4-gb300-dynamo-sglang:
-  image: lmsysorg/sglang:v0.5.14-cu130@sha256:5027e95bf6ec536856b1b52a91d1f35ff5c564ab83e8a94758a169ff09bb8df3
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: gb300
-  precision: fp4
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "81d0555ee23519cea80a42b4fe824e30368b7300" }
-  kv-p2p-transfer: mooncake
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # WideEP TP=16 decode: 1p1d-dep4-dep16. 5 nodes (4P + 16D = 20 GPUs).
-      - conc-list: [1024]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb300-fp4/8k1k/disagg-1p1d-dep4-dep16-5-c1024-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # Low concurrency: 1p1d-tp4-tp4. 2 nodes.
-      - conc-list: [1]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb300-fp4/8k1k/disagg-1p1d-tp4-tp4-2-c1-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # --- Weiliang wide-EP sweep (srt-slurm PR#173), 18 nodes total ---
-      # EP=12: 15P+3D, conc=12000.
-      - conc-list: [12000]
-        prefill:
-          num-worker: 15
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb300-fp4/8k1k/disagg-15p1d-dep4-dep12-18-c12000-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 12
-          ep: 12
-          dp-attn: true
-      # EP=16: 14P+4D, conc=8192.
-      - conc-list: [8192]
-        prefill:
-          num-worker: 14
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb300-fp4/8k1k/disagg-14p1d-dep4-dep16-18-c8192-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # EP=24: 12P+6D, conc=3000.
-      - conc-list: [3000]
-        prefill:
-          num-worker: 12
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb300-fp4/8k1k/disagg-12p1d-dep4-dep24-18-c3000-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 24
-          ep: 24
-          dp-attn: true
-      # EP=32: 10P+8D, conc=2500.
-      - conc-list: [2500]
-        prefill:
-          num-worker: 10
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb300-fp4/8k1k/disagg-10p1d-dep4-dep32-18-c2500-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      # EP=40: 8P+10D, conc=2048.
-      - conc-list: [2048]
-        prefill:
-          num-worker: 8
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb300-fp4/8k1k/disagg-8p1d-dep4-dep40-18-c2048-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 40
-          ep: 40
-          dp-attn: true
-
-# MTP variant of dsv4-fp4-gb300-dynamo-sglang.
-dsv4-fp4-gb300-dynamo-sglang-mtp:
-  image: lmsysorg/sglang:nightly-dev-20260527-14f81a67
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: gb300
-  precision: fp4
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "81d0555ee23519cea80a42b4fe824e30368b7300" }
-  kv-p2p-transfer: mooncake
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # Low-latency baseline: 1p1d-tp4-tp4. 2 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [1]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb300-fp4/8k1k/disagg-1p1d-p-tp4-d-tp4-b8-c1-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # Low-latency 1p6d-dep4-tp4: 1P (DEP=4) + 6 TP=4 decode workers. 7 nodes.
-      # Recipe runs concurrencies=8x32x64; matrix tracks the max.
-      - spec-decoding: "mtp"
-        conc-list: [64]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb300-fp4/8k1k/disagg-1p6d-p-tp4-ep4-dp4-d-tp4-b128-c8x32x64-mtp.yaml"
-        decode:
-          num-worker: 6
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # Mid curve 1p1d-dep4-dep8. 3 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [256]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb300-fp4/8k1k/disagg-1p1d-p-tp4-ep4-dp4-d-tp8-ep8-dp8-b3072-c256-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # Mid curve 1p1d-dep4-dep16. 5 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [256]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb300-fp4/8k1k/disagg-1p1d-p-tp4-ep4-dp4-d-tp16-ep16-dp16-b3072-c256-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # Mid curve 2p1d-dep4-dep8. 4 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [512]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb300-fp4/8k1k/disagg-2p1d-p-tp4-ep4-dp4-d-tp8-ep8-dp8-b3072-c512-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # Mid curve 4p1d-dep4-dep8. 6 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [1024]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb300-fp4/8k1k/disagg-4p1d-p-tp4-ep4-dp4-d-tp8-ep8-dp8-b3072-c1024-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # High concurrency 6p1d-dep4-dep8. 8 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [4096]
-        prefill:
-          num-worker: 6
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb300-fp4/8k1k/disagg-high-conc-6p1d-dep4-dep8-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # High concurrency 8p1d-dep4-dep8. 10 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [8192]
-        prefill:
-          num-worker: 8
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/gb300-fp4/8k1k/disagg-high-conc-8p1d-dep4-dep8-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-# DSV4 B300 disaggregated dynamo-sglang MTP configuration.
-# Each search-space entry maps one-to-one to a checked-in srt-slurm recipe under
-# recipes/sglang/deepseek-v4/8k1k/. Runs on the b300 pool via
-# runners/launch_b300-nv.sh. Prefill workers are TP4/DEP4 (two workers per
-# 8-GPU B300 node); decode is DEP8 (one full node) except the TP4 topologies.
-# EAGLE draft depth is tapered down as concurrency grows
-# (3 steps / 4 draft tokens -> 2/3 at 6p1d -> 1/2 at 8p1d).
-dsv4-fp4-b300-dynamo-sglang-mtp:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260721-8905cbd4
-  model: deepseek-ai/DeepSeek-V4-Pro
-  model-prefix: dsv4
-  runner: b300
-  precision: fp4
-  framework: dynamo-sglang
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 1p1d-tp4-tp4 (pure TP, no dp-attn). 2 nodes (1P + 1D).
-      - spec-decoding: "mtp"
-        conc-list: [1]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b300-fp4/8k1k/disagg-1p1d-tp4-tp4-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # Scale-out: 1x DEP4 prefill + 6x TP4 decode. 4 nodes (1P + 3D).
-      - spec-decoding: "mtp"
-        conc-list: [8, 32, 64]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b300-fp4/8k1k/disagg-1p6d-dep4-tp4-mtp.yaml"
-        decode:
-          num-worker: 6
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p1d-dep4-dep8: 1x DEP4 prefill + DEP8 decode. 2 nodes (1P + 1D).
-      - spec-decoding: "mtp"
-        conc-list: [256]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b300-fp4/8k1k/disagg-1p1d-dep4-dep8-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # 1p2d-dep4-dep8: 1x DEP4 prefill + 2x DEP8 decode. 3 nodes (1P + 2D).
-      - spec-decoding: "mtp"
-        conc-list: [256]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b300-fp4/8k1k/disagg-1p2d-dep4-dep8-mtp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # 2p1d-dep4-dep8: 2x DEP4 prefill + DEP8 decode. 2 nodes (1P + 1D).
-      - spec-decoding: "mtp"
-        conc-list: [512]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b300-fp4/8k1k/disagg-2p1d-dep4-dep8-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # 4p1d-dep4-dep8: 4x DEP4 prefill + DEP8 decode. 3 nodes (2P + 1D).
-      - spec-decoding: "mtp"
-        conc-list: [1024]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b300-fp4/8k1k/disagg-4p1d-dep4-dep8-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # 6p1d-dep4-dep8: 6x DEP4 prefill + DEP8 decode. 4 nodes (3P + 1D).
-      - spec-decoding: "mtp"
-        conc-list: [2048]
-        prefill:
-          num-worker: 6
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b300-fp4/8k1k/disagg-6p1d-dep4-dep8-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      # 8p1d-dep4-dep8. 5 nodes (4P + 1D).
-      - spec-decoding: "mtp"
-        conc-list: [3072]
-        prefill:
-          num-worker: 8
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/dsv4/sglang/b300-fp4/8k1k/disagg-8p1d-dep4-dep8-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-# Archive source: nvidia-glm5-glm5.1-master.yaml
-# Deprecated GLM-5 / GLM-5.1 entries archived from nvidia-master.yaml.
-# Removed from the active master config so sweep generation no longer selects them.
-
-glm5-fp8-b200-sglang:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260605-7dc73766
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: b200
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 256 }
-
-
-glm5-fp8-b200-sglang-mtp:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260605-7dc73766
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: b200
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 256, spec-decoding: mtp }
-
-  # NOTE: At the time of submission, https://cookbook.sglang.io/autoregressive/GLM/GLM-5.1
-  # does not have a B300-specific recipe, so this config reuses the existing GLM5 FP8
-  # B200 SGLang recipe as-is until B300-specific tuning is available.
-
-
-glm5-fp8-b300-sglang:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: b300
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 256 }
-
-
-glm5-fp8-b300-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: b300
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 256, spec-decoding: mtp }
-
-
-glm5-fp4-b200-sglang:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260605-7dc73766
-  model: nvidia/GLM-5-NVFP4
-  model-prefix: glm5
-  runner: b200
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 4 }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 }
-
-
-glm5-fp4-b200-sglang-mtp:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260605-7dc73766
-  model: nvidia/GLM-5-NVFP4
-  model-prefix: glm5
-  runner: b200
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 4, spec-decoding: mtp }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 256, spec-decoding: mtp }
-
-
-  # NOTE: At the time of submission, https://cookbook.sglang.io/autoregressive/GLM/GLM-5
-  # does not have a B300-specific recipe, so this config reuses the existing
-  # GLM-5 FP4 B200 SGLang recipe as-is until B300-specific tuning is available.
-
-glm5-fp4-b300-sglang:
-  image: lmsysorg/sglang:v0.5.11-cu130
-  model: nvidia/GLM-5-NVFP4
-  model-prefix: glm5
-  runner: b300
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 4 }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 256 }
-
-
-glm5-fp4-b300-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.11-cu130
-  model: nvidia/GLM-5-NVFP4
-  model-prefix: glm5
-  runner: b300
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 4, spec-decoding: mtp }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 256, spec-decoding: mtp }
-
-
-glm5-fp4-gb300-dynamo-trt:
-  image: nvcr.io/nvidia/ai-dynamo/tensorrtllm-runtime:1.3.0-dev.1-cuda13
-  model: nvidia/GLM-5-NVFP4
-  model-prefix: glm5
-  runner: gb300-nv
-  precision: fp4
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "v1.3.0-dev.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # STP configurations
-      - conc-list: [ 10 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/STP/ctx1dep2_gen5tep4_batch2_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/STP/ctx1dep2_gen5tep4_batch2_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 25 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/STP/ctx1dep2_gen5tep4_batch4_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/STP/ctx1dep2_gen5tep4_batch4_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 50 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/STP/ctx1dep2_gen5tep4_batch8_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/STP/ctx1dep2_gen5tep4_batch8_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 100 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/STP/ctx1dep2_gen5tep4_batch16_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/STP/ctx1dep2_gen5tep4_batch16_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 308 ]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/STP/ctx3dep2_gen1dep32_batch8_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/STP/ctx3dep2_gen1dep32_batch8_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 615 ]
-        prefill:
-          num-worker: 6
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/STP/ctx6dep2_gen1dep32_batch16_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/STP/ctx6dep2_gen1dep32_batch16_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 1076 ]
-        prefill:
-          num-worker: 9
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/STP/ctx9dep2_gen1dep16_batch64_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/STP/ctx9dep2_gen1dep16_batch64_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [ 1229 ]
-        prefill:
-          num-worker: 11
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/STP/ctx11dep2_gen1dep32_batch32_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/STP/ctx11dep2_gen1dep32_batch32_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 2151 ]
-        prefill:
-          num-worker: 15
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/STP/ctx15dep2_gen1dep16_batch128_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/STP/ctx15dep2_gen1dep16_batch128_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-
-glm5-fp4-gb200-dynamo-trt:
-  image: nvcr.io/nvidia/ai-dynamo/tensorrtllm-runtime:1.3.0-dev.1-cuda13
-  model: nvidia/GLM-5-NVFP4
-  model-prefix: glm5
-  runner: gb200
-  precision: fp4
-  framework: dynamo-trt
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # STP configurations
-      - conc-list: [ 5 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-1p5d-dep4-tep4-b1-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 10 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch2_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-1p5d-dep4-tep4-b2-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 25 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch4_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-1p5d-dep4-tep4-b4-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 50 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch8_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-1p5d-dep4-tep4-b8-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 105 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch16_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-1p5d-dep4-tep4-b16-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 308 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/STP/ctx2dep4_gen1dep32_batch8_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-2p1d-dep4-dep32-b8-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 615 ]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/STP/ctx4dep4_gen1dep32_batch16_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-4p1d-dep4-dep32-b16-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 1127 ]
-        prefill:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/STP/ctx5dep4_gen1dep16_batch64_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-5p1d-dep4-dep16-b64-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [ 1229 ]
-        prefill:
-          num-worker: 6
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/STP/ctx6dep4_gen1dep32_batch32_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-6p1d-dep4-dep32-b32-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 2151 ]
-        prefill:
-          num-worker: 9
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/STP/ctx9dep4_gen1dep16_batch128_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-9p1d-dep4-dep16-b128-eplb0-mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-
-glm5-fp4-gb200-dynamo-trt-mtp:
-  image: nvcr.io/nvidia/ai-dynamo/tensorrtllm-runtime:1.3.0-dev.1-cuda13
-  model: nvidia/GLM-5-NVFP4
-  model-prefix: glm5
-  runner: gb200
-  precision: fp4
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "v1.3.0-dev.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # MTP configurations
-      - spec-decoding: "mtp"
-        conc-list: [ 5 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/MTP/ctx1dep4_gen5tep4_batch1_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-1p5d-dep4-tep4-b1-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [ 15 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/MTP/ctx1dep4_gen5tep4_batch2_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-1p5d-dep4-tep4-b2-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [ 30 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/MTP/ctx1dep4_gen5tep4_batch4_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-1p5d-dep4-tep4-b4-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [ 90 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/MTP/ctx1dep4_gen1dep32_batch2_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-1p1d-dep4-dep32-b2-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 180 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/MTP/ctx2dep4_gen1dep32_batch4_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-2p1d-dep4-dep32-b4-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 333 ]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/MTP/ctx4dep4_gen1dep32_batch8_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-4p1d-dep4-dep32-b8-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 615 ]
-        prefill:
-          num-worker: 6
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/MTP/ctx6dep4_gen1dep16_batch32_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-6p1d-dep4-dep16-b32-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 1127 ]
-        prefill:
-          num-worker: 8
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/MTP/ctx8dep4_gen1dep16_batch64_eplb0_mtp1.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-8p1d-dep4-dep16-b64-eplb0-mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 666 ]
-        prefill:
-          num-worker: 7
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/MTP/ctx7dep4_gen1dep32_batch16_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-7p1d-dep4-dep32-b16-eplb0-mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 2253 ]
-        prefill:
-          num-worker: 12
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/v1.0.26/recipes/trtllm/glm5/disagg/trtllm_dynamo/gb200_nvfp4/ISL8K_OSL1K/MTP/ctx12dep4_gen1dep16_batch128_eplb0_mtp1.yaml
-          - "CONFIG_FILE=recipes/glm5/trtllm/gb200-fp4/8k1k/disagg-12p1d-dep4-dep16-b128-eplb0-mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-
-glm5-fp8-h200-sglang:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: h200-dgxc
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-
-glm5-fp8-h200-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: h200
-  precision: fp8
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-
-
-glm5-fp4-gb200-dynamo-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.13.post1-cu130
-  model: nvidia/GLM-5.1-NVFP4
-  model-prefix: glm5.1
-  runner: gb200
-  precision: fp4
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "1.2.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 1p1d dep16. 5 nodes (1P + 4D).
-      - spec-decoding: "mtp"
-        conc-list: [109]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_8k1k_hightpt[0]"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # 2p1d dep16. 6 nodes (2P + 4D).
-      - spec-decoding: "mtp"
-        conc-list: [398]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_8k1k_hightpt[1]"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # 4p1d dep16. 8 nodes (4P + 4D).
-      - spec-decoding: "mtp"
-        conc-list: [853]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_8k1k_hightpt[2]"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # 8p1d dep16. 12 nodes (8P + 4D).
-      - spec-decoding: "mtp"
-        conc-list: [2132]
-        prefill:
-          num-worker: 8
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_8k1k_hightpt[3]"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-    # ---------- 8k1k low-latency (per-node TP=4 decode workers, MTP) ----------
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 1p4d tp4. 5 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [104]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_8k1k_lowlat[0]"
-        decode:
-          num-worker: 4
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p4d tp4, conc16. 5 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [82]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_8k1k_lowlat[1]"
-        decode:
-          num-worker: 4
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p8d tp4. 9 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [87]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_8k1k_lowlat[2]"
-        decode:
-          num-worker: 8
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p8d tp4, conc4. 9 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [47]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_8k1k_lowlat[3]"
-        decode:
-          num-worker: 8
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p16d tp4. 17 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [74]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_8k1k_lowlat[4]"
-        decode:
-          num-worker: 16
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p16d tp4, conc1. 17 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [26]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_8k1k_lowlat[5]"
-        decode:
-          num-worker: 16
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-
-glm5-fp8-b200-dynamo-sglang:
-  image: lmsysorg/sglang:v0.5.11-cu130
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: b200-dgxc
-  precision: fp8
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "0.8.0" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [560]
-        prefill:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/glm5.yaml:zip_override_8k1k_hightpt[0]"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-      - conc-list: [240]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/glm5.yaml:zip_override_8k1k_hightpt[1]"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-      - conc-list: [224]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/glm5.yaml:zip_override_8k1k_hightpt[2]"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: true
-      - conc-list: [256]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/glm5.yaml:zip_override_8k1k_lowlat[0]"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [256]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/glm5.yaml:zip_override_8k1k_lowlat[1]"
-        decode:
-          num-worker: 3
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [200]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/glm5.yaml:zip_override_8k1k_lowlat[2]"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [128]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/glm5.yaml:zip_override_8k1k_lowlat[3]"
-        decode:
-          num-worker: 5
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [64]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/glm5.yaml:zip_override_8k1k_lowlat[4]"
-        decode:
-          num-worker: 7
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [12]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/b200-fp8/glm5.yaml:zip_override_8k1k_lowlat[5]"
-        decode:
-          num-worker: 8
-          tp: 8
-          ep: 1
-          dp-attn: false
-
-
-glm5-fp4-gb200-dynamo-sglang:
-  image: lmsysorg/sglang:v0.5.11-cu130
-  model: nvidia/GLM-5.1-NVFP4
-  model-prefix: glm5.1
-  runner: gb200
-  precision: fp4
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "1.2.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 4p1d wide-EP. 12 nodes.
-      - conc-list: [989]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/8k1k/disagg-4p1d-p-tp4-ep4-dp4-d-tp32-ep32-dp32-b599-c989-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      # 4p1d wide-EP, mrr512. 12 nodes.
-      - conc-list: [686]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/8k1k/disagg-4p1d-p-tp4-ep4-dp4-d-tp32-ep32-dp32-b512-c686-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      # 6p1d wide-EP. 14 nodes.
-      - conc-list: [1497]
-        prefill:
-          num-worker: 6
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/8k1k/disagg-6p1d-p-tp4-ep4-dp4-d-tp32-ep32-dp32-b914-c1497-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      # 10p1d wide-EP. 18 nodes.
-      - conc-list: [2674]
-        prefill:
-          num-worker: 10
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/8k1k/disagg-10p1d-p-tp4-ep4-dp4-d-tp32-ep32-dp32-b1712-c2674-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-    # ---------- 8k1k low-latency (per-node TP=4 decode workers) ----------
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 1p4d. 5 nodes.
-      - conc-list: [133]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/8k1k/disagg-1p4d-p-tp4-ep4-dp4-d-tp4-b32-c133-stp.yaml"
-        decode:
-          num-worker: 4
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p6d. 7 nodes.
-      - conc-list: [146]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/8k1k/disagg-1p6d-p-tp4-ep4-dp4-d-tp4-b22-c146-stp.yaml"
-        decode:
-          num-worker: 6
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p6d, mrr16. 7 nodes.
-      - conc-list: [103]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/8k1k/disagg-1p6d-p-tp4-ep4-dp4-d-tp4-b16-c103-stp.yaml"
-        decode:
-          num-worker: 6
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p8d. 9 nodes.
-      - conc-list: [130]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/8k1k/disagg-1p8d-p-tp4-ep4-dp4-d-tp4-b15-c130-stp.yaml"
-        decode:
-          num-worker: 8
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p16d. 17 nodes.
-      - conc-list: [113]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/8k1k/disagg-1p16d-p-tp4-ep4-dp4-d-tp4-b6-c113-stp.yaml"
-        decode:
-          num-worker: 16
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p16d, mrr1 (single-stream). 17 nodes.
-      - conc-list: [23]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp4/8k1k/disagg-1p16d-p-tp4-ep4-dp4-d-tp4-b1-c23-stp.yaml"
-        decode:
-          num-worker: 16
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-
-glm5-fp4-gb300-dynamo-sglang:
-  image: lmsysorg/sglang:v0.5.11-cu130
-  model: nvidia/GLM-5.1-NVFP4
-  model-prefix: glm5
-  runner: gb300-nv
-  precision: fp4
-  framework: dynamo-sglang
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 5p1d wide-EP. 13 nodes (5P @ TP=4 + 1D @ TP=32 on 8 nodes).
-      - conc-list: [2048]
-        prefill:
-          num-worker: 5
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5.yaml:zip_override_8k1k_hightpt[0]"
-        router: { name: dynamo-router, version: "0.8.0" }
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      # 7p1d wide-EP. 15 nodes.
-      - conc-list: [3072]
-        prefill:
-          num-worker: 7
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5.yaml:zip_override_8k1k_hightpt[1]"
-        router: { name: dynamo-router, version: "0.8.0" }
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      # 10p1d wide-EP. 18 nodes.
-      - conc-list: [4096]
-        prefill:
-          num-worker: 10
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5.yaml:zip_override_8k1k_hightpt[2]"
-        router: { name: dynamo-router, version: "0.8.0" }
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-    # ---------- 8k1k low-latency (per-node TP=4 decode workers) ----------
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 1p3d. 4 nodes (1P + 3 D workers @ 1 node each).
-      - conc-list: [128]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5/sglang/gb300-fp4/8k1k/disagg-1p3d-p-tp4-dp4-d-tp4-b128-c128-stp.yaml"
-        router: { name: dynamo-router, version: "1.1.0" }
-        decode:
-          num-worker: 3
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p5d. 6 nodes.
-      - conc-list: [64]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5/sglang/gb300-fp4/8k1k/disagg-1p5d-p-tp4-dp4-d-tp4-b64-c64-stp.yaml"
-        router: { name: dynamo-router, version: "1.1.0" }
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p9d. 10 nodes.
-      - conc-list: [32]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5/sglang/gb300-fp4/8k1k/disagg-1p9d-p-tp4-dp4-d-tp4-b32-c32-stp.yaml"
-        router: { name: dynamo-router, version: "1.1.0" }
-        decode:
-          num-worker: 9
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p15d. 16 nodes.
-      - conc-list: [16]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5/sglang/gb300-fp4/8k1k/disagg-1p15d-p-tp4-dp4-d-tp4-b16-c16-stp.yaml"
-        router: { name: dynamo-router, version: "1.1.0" }
-        decode:
-          num-worker: 15
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p17d. 18 nodes.
-      - conc-list: [12]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5/sglang/gb300-fp4/8k1k/disagg-1p17d-p-tp4-dp4-d-tp4-b1-c12-stp.yaml"
-        router: { name: dynamo-router, version: "1.1.0" }
-        decode:
-          num-worker: 17
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-
-    # ---------- 1k1k high-throughput (wide-EP TP=32 decode) ----------
-
-glm5-fp8-gb200-dynamo-sglang:
-  image: lmsysorg/sglang:v0.5.12
-  model: zai-org/GLM-5.1-FP8
-  model-prefix: glm5.1
-  runner: gb200
-  precision: fp8
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "1.2.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 3p1d wide-EP (dep32, mrr256). 14 nodes.
-      - conc-list: [519]
-        prefill:
-          num-worker: 3
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/8k1k/disagg-3p1d-p-tp8-dp8-d-tp32-ep32-dp32-b256-c519-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      # 4p1d wide-EP (dep16). 12 nodes.
-      - conc-list: [1484]
-        prefill:
-          num-worker: 4
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/8k1k/disagg-4p1d-p-tp8-dp8-d-tp16-ep16-dp16-b802-c1484-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # 5p1d wide-EP (dep32). 18 nodes.
-      - conc-list: [1688]
-        prefill:
-          num-worker: 5
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/8k1k/disagg-5p1d-p-tp8-dp8-d-tp32-ep32-dp32-b828-c1688-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      # 7p1d wide-EP (dep16). 18 nodes.
-      - conc-list: [2699]
-        prefill:
-          num-worker: 7
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/8k1k/disagg-7p1d-p-tp8-dp8-d-tp16-ep16-dp16-b1514-c2699-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-    # ---------- 8k1k low-latency (per-node TP=8 decode workers) ----------
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 1p8d, mrr1 (single-stream). 18 nodes.
-      - conc-list: [8]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/8k1k/disagg-1p8d-p-tp8-dp8-d-tp8-b1-c8-stp.yaml"
-        decode:
-          num-worker: 8
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # 1p8d, mrr16. 18 nodes.
-      - conc-list: [90]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/8k1k/disagg-1p8d-p-tp8-dp8-d-tp8-b16-c90-stp.yaml"
-        decode:
-          num-worker: 8
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # 1p8d, mrr4. 18 nodes.
-      - conc-list: [9]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/8k1k/disagg-1p8d-p-tp8-dp8-d-tp8-b4-c9-stp.yaml"
-        decode:
-          num-worker: 8
-          tp: 8
-          ep: 1
-          dp-attn: false
-
-
-    # ---------- 1k1k high-throughput (wide-EP decode) ----------
-
-glm5-fp8-gb200-dynamo-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.13.post1-cu130
-  model: zai-org/GLM-5.1-FP8
-  model-prefix: glm5.1
-  runner: gb200
-  precision: fp8
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "1.2.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    # ---------- 8k1k high-throughput (wide-EP TP=16 decode, MTP) ----------
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 1p1d dep16. 6 nodes (2P + 4D).
-      - spec-decoding: "mtp"
-        conc-list: [137]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_8k1k_hightpt[0]"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # 2p1d dep16. 8 nodes (4P + 4D).
-      - spec-decoding: "mtp"
-        conc-list: [620]
-        prefill:
-          num-worker: 2
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_8k1k_hightpt[1]"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # 4p1d dep16. 12 nodes (8P + 4D).
-      - spec-decoding: "mtp"
-        conc-list: [1305]
-        prefill:
-          num-worker: 4
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_8k1k_hightpt[2]"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-    # ---------- 8k1k low-latency (TP=8 decode workers, MTP) ----------
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 1p4d tp8. 10 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [136]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_8k1k_lowlat[0]"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # 1p4d tp8, conc56. 10 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [56]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_8k1k_lowlat[1]"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # 1p8d tp8. 18 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [120]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_8k1k_lowlat[2]"
-        decode:
-          num-worker: 8
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # 1p8d tp8, conc12. 18 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [12]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_8k1k_lowlat[3]"
-        decode:
-          num-worker: 8
-          tp: 8
-          ep: 1
-          dp-attn: false
-      # 1p8d tp8, conc31. 18 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [31]
-        prefill:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb200-fp8/fixed-seq-len/disagg-mtp-variants.yaml:zip_override_mtp_8k1k_lowlat[4]"
-        decode:
-          num-worker: 8
-          tp: 8
-          ep: 1
-          dp-attn: false
-
-
-glm5-fp4-gb300-dynamo-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.13.post1-cu130
-  model: nvidia/GLM-5.1-NVFP4
-  model-prefix: glm5.1
-  runner: gb300-nv
-  precision: fp4
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "0.8.0" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 4p1d dep16. 6 nodes (2P + 4D).
-      - spec-decoding: "mtp"
-        conc-list: [364]
-        prefill:
-          num-worker: 4
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5-mtp.yaml:zip_override_mtp_8k1k_hightpt[0]"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      # 8p1d dep16. 8 nodes (4P + 4D).
-      - spec-decoding: "mtp"
-        conc-list: [512]
-        prefill:
-          num-worker: 8
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5-mtp.yaml:zip_override_mtp_8k1k_hightpt[1]"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-    # ---------- 8k1k low-latency (per-node TP=4 decode workers, MTP) ----------
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 1p2d tp4. 3 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [66]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5-mtp.yaml:zip_override_mtp_8k1k_lowlat[0]"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p2d tp4, conc16. 3 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [41]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5-mtp.yaml:zip_override_mtp_8k1k_lowlat[1]"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p4d tp4. 5 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [55]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5-mtp.yaml:zip_override_mtp_8k1k_lowlat[2]"
-        decode:
-          num-worker: 4
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p4d tp4, conc4. 5 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [23]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5-mtp.yaml:zip_override_mtp_8k1k_lowlat[3]"
-        decode:
-          num-worker: 4
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p8d tp4. 9 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [44]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5-mtp.yaml:zip_override_mtp_8k1k_lowlat[4]"
-        decode:
-          num-worker: 8
-          tp: 4
-          ep: 1
-          dp-attn: false
-      # 1p8d tp4, conc1. 9 nodes.
-      - spec-decoding: "mtp"
-        conc-list: [12]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp4/glm5-mtp.yaml:zip_override_mtp_8k1k_lowlat[5]"
-        decode:
-          num-worker: 8
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-
-    # ---------- 1k1k high-throughput (wide-EP decode, MTP) ----------
-
-glm5-fp8-gb300-dynamo-sglang:
-  image: lmsysorg/sglang:v0.5.11-cu130
-  model: zai-org/GLM-5-FP8
-  model-prefix: glm5
-  runner: gb300-nv
-  precision: fp8
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "0.8.0" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [2800]
-        prefill:
-          num-worker: 14
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp8/glm5.yaml:zip_override_8k1k_hightpt[0]"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [1700]
-        prefill:
-          num-worker: 12
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp8/glm5.yaml:zip_override_8k1k_hightpt[1]"
-        decode:
-          num-worker: 1
-          tp: 24
-          ep: 24
-          dp-attn: true
-      - conc-list: [1300]
-        prefill:
-          num-worker: 10
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp8/glm5.yaml:zip_override_8k1k_hightpt[2]"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [900]
-        prefill:
-          num-worker: 8
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp8/glm5.yaml:zip_override_8k1k_hightpt[3]"
-        decode:
-          num-worker: 1
-          tp: 40
-          ep: 40
-          dp-attn: true
-    # ---------- 8k1k low-latency (per-node TP=4 decode workers) ----------
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [150]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp8/glm5.yaml:zip_override_8k1k_lowlat[0]"
-        decode:
-          num-worker: 9
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [128, 64, 32]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp8/glm5.yaml:zip_override_8k1k_lowlat[1]"
-        decode:
-          num-worker: 17
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [24]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/gb300-fp8/glm5.yaml:zip_override_8k1k_lowlat[2]"
-        decode:
-          num-worker: 17
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-
-    # ---------- 1k1k high-throughput (wide-EP decode) ----------
-
-glm5-fp8-gb300-dynamo-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.11-cu130
-  model: zai-org/GLM-5.1-FP8
-  model-prefix: glm5.1
-  runner: gb300-nv
-  precision: fp8
-  framework: dynamo-sglang
-  router: { name: dynamo-router, version: "1.2.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - spec-decoding: "mtp"
-        conc-list: [2800]
-        prefill:
-          num-worker: 14
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb300-fp8/8k1k/disagg-14p1d-p-tp4-dp4-d-tp16-ep16-dp16-b2800-c2800-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [1700]
-        prefill:
-          num-worker: 12
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb300-fp8/8k1k/disagg-12p1d-p-tp4-dp4-d-tp24-ep24-dp24-b1680-c1700-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 24
-          ep: 24
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [1300]
-        prefill:
-          num-worker: 10
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb300-fp8/8k1k/disagg-10p1d-p-tp4-dp4-d-tp32-ep32-dp32-b1280-c1300-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [900]
-        prefill:
-          num-worker: 8
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb300-fp8/8k1k/disagg-8p1d-p-tp4-dp4-d-tp40-ep40-dp40-b880-c900-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 40
-          ep: 40
-          dp-attn: true
-    # ---------- 8k1k low-latency (per-node TP=4 decode workers, EAGLE MTP) ----------
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - spec-decoding: "mtp"
-        conc-list: [150]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb300-fp8/8k1k/disagg-1p9d-p-tp4-dp4-d-tp4-b15-c150-mtp.yaml"
-        decode:
-          num-worker: 9
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [128, 64, 32]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb300-fp8/8k1k/disagg-1p17d-p-tp4-dp4-d-tp4-b8-c128x64x32-mtp.yaml"
-        decode:
-          num-worker: 17
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [24]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/glm5.1/sglang/gb300-fp8/8k1k/disagg-1p17d-p-tp4-dp4-d-tp4-b1-c24-mtp.yaml"
-        decode:
-          num-worker: 17
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-
-glm5-fp4-gb300-dynamo-trt-mtp:
-  image: nvcr.io/nvidia/ai-dynamo/tensorrtllm-runtime:1.3.0-dev.1-cuda13
-  model: nvidia/GLM-5-NVFP4
-  model-prefix: glm5
-  runner: gb300-nv
-  precision: fp4
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "v1.3.0-dev.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # MTP configurations
-      - spec-decoding: "mtp"
-        conc-list: [ 5 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx1dep2_gen5tep4_batch1_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx1dep2_gen5tep4_batch1_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [ 15 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx1dep2_gen5tep4_batch2_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx1dep2_gen5tep4_batch2_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [ 30 ]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx1dep2_gen5tep4_batch4_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx1dep2_gen5tep4_batch4_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [ 84 ]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx2dep2_gen1dep32_batch2_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx2dep2_gen1dep32_batch2_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 180 ]
-        prefill:
-          num-worker: 4
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx4dep2_gen1dep32_batch4_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx4dep2_gen1dep32_batch4_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 333 ]
-        prefill:
-          num-worker: 6
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx6dep2_gen1dep32_batch8_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx6dep2_gen1dep32_batch8_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 615 ]
-        prefill:
-          num-worker: 10
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx10dep2_gen1dep16_batch32_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx10dep2_gen1dep16_batch32_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 1229 ]
-        prefill:
-          num-worker: 14
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx14dep2_gen1dep16_batch64_eplb0_mtp1.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx14dep2_gen1dep16_batch64_eplb0_mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 615 ]
-        prefill:
-          num-worker: 11
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx11dep2_gen1dep32_batch16_eplb0_mtp3.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx11dep2_gen1dep32_batch16_eplb0_mtp3.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [ 2253 ]
-        prefill:
-          num-worker: 21
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx21dep2_gen1dep16_batch128_eplb0_mtp1.yaml
-          - "CONFIG_FILE=recipes/GLM5/disagg/trtllm_dynamo/gb300_nvfp4/ISL8K_OSL1K/MTP/ctx21dep2_gen1dep16_batch128_eplb0_mtp1.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-# Archive source: nvidia-gptoss-master.yaml
-# Deprecated gpt-oss-120b entries archived from nvidia-master.yaml.
-# Removed from the active master config so sweep generation no longer selects them.
-
-gptoss-fp4-b200-trt:
-  image: nvcr.io#nvidia/tensorrt-llm/release:1.3.0rc14
-  model: openai/gpt-oss-120b
-  model-prefix: gptoss
-  runner: b200
-  precision: fp4
-  framework: trt
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    # Low ==> high TP from Left to Right of pareto
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 256, conc-end: 256 }
-      - { tp: 2, ep: 2, dp-attn: true, conc-start: 256, conc-end: 256 }
-      - { tp: 2, conc-start:  4, conc-end: 256 }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 64, conc-end: 64 }
-      - { tp: 4, conc-start: 4, conc-end: 4 }
-      - { tp: 8, conc-start: 4, conc-end: 4 }
-    # Low ==> high TP from Left to Right of pareto
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start:   4, conc-end: 256}
-      - { tp: 2, conc-start:   4, conc-end: 256}
-      - { tp: 4, conc-start:   4, conc-end:  32}
-      - { tp: 8, conc-start:   4, conc-end:   4}
-
-gptoss-fp4-b200-vllm:
-  image: vllm/vllm-openai:v0.22.0
-  model: openai/gpt-oss-120b
-  model-prefix: gptoss
-  runner: b200
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 4, conc-end: 128 }
-      - { tp: 2, conc-start: 4, conc-end: 128 }
-      - { tp: 4, conc-start: 4, conc-end: 64 }
-      - { tp: 8, conc-start: 4, conc-end: 8 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 4, conc-end: 128 }
-      - { tp: 2, conc-start: 4, conc-end: 128 }
-      - { tp: 4, conc-start: 4, conc-end: 64 }
-      - { tp: 8, conc-start: 4, conc-end: 4 }
-
-gptoss-fp4-h100-vllm:
-  image: vllm/vllm-openai:v0.21.0
-  model: openai/gpt-oss-120b
-  model-prefix: gptoss
-  runner: h100
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 2, conc-start: 4, conc-end: 64 }
-      - { tp: 4, conc-start: 4, conc-end: 64 }
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 2, conc-start: 4, conc-end: 64 }
-      - { tp: 4, conc-start: 4, conc-end: 64 }
-      - { tp: 8, conc-start: 4, conc-end: 16 }
-
-# Day-zero MiniMax-M3 recipe (https://recipes.vllm.ai/MiniMaxAI/MiniMax-M3).
-# M3 support has not shipped in a stable vLLM release; the dedicated
-# vllm/vllm-openai:minimax-m3 image is the supported path. MXFP8 variant
-# (NVIDIA-quantized, ~427 GB weights) is the lowest precision available —
-# BF16 (~854 GB) does not fit 8x H100 (640 GB) at all, so H100 is TP8-only:
-# weights alone take ~56 GB of each 80 GB GPU, leaving no room below TP8.
-# dp-attn: true maps to the recipe's "DP8 + Expert Parallel" serve mode
-# (vLLM --data-parallel-size 8 --enable-expert-parallel).
-
-gptoss-fp4-h200-trt:
-  image: nvcr.io#nvidia/tensorrt-llm/release:1.3.0rc14
-  model: openai/gpt-oss-120b
-  model-prefix: gptoss
-  runner: h200
-  precision: fp4
-  framework: trt
-  multinode: false
-  # For all sequence lengths, EP=TP, DP_ATTENTION=false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 1, ep: 1, dp-attn: false, conc-start: 4, conc-end: 64 }
-      - { tp: 2, ep: 2, dp-attn: false, conc-start: 4, conc-end: 64 }
-      - { tp: 4, ep: 4, dp-attn: false, conc-start: 4, conc-end: 32 }
-      - { tp: 8, ep: 8, dp-attn: false, conc-start: 4, conc-end: 8 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 1, ep: 1, dp-attn: false, conc-start: 4, conc-end: 64 }
-      - { tp: 2, ep: 2, dp-attn: false, conc-start: 4, conc-end: 64 }
-      - { tp: 4, ep: 4, dp-attn: false, conc-start: 4, conc-end: 64 }
-      - { tp: 8, ep: 8, dp-attn: false, conc-start: 4, conc-end: 8 }
-
-gptoss-fp4-h200-vllm:
-  image: vllm/vllm-openai:v0.22.0
-  model: openai/gpt-oss-120b
-  model-prefix: gptoss
-  runner: h200
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 4, conc-end: 4 }
-      - { tp: 2, conc-start: 4, conc-end: 64 }
-      - { tp: 4, conc-start: 4, conc-end: 64 }
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 4, conc-end: 64 }
-      - { tp: 2, conc-start: 4, conc-end: 64 }
-      - { tp: 4, conc-start: 4, conc-end: 64 }
-      - { tp: 8, conc-start: 4, conc-end: 32 }
-
-# Day-zero MiniMax-M3 recipe (https://recipes.vllm.ai/MiniMaxAI/MiniMax-M3).
-# Dedicated vllm/vllm-openai:minimax-m3 image (no stable release has M3 yet).
-# MXFP8 variant (~427 GB weights) is the lowest precision available; on
-# 8x H200 (1128 GB) it leaves ample KV headroom where BF16 is a tight fit.
-# TP4 (~112 GB weights/GPU) is memory-tight — swept only at low/mid conc.
-# dp-attn: true maps to the recipe's "DP8 + Expert Parallel" serve mode.
-
-gptoss-fp4-gb200-dynamo-trt:
-  image: nvcr.io#nvidia/ai-dynamo/tensorrtllm-runtime:0.7.0.post2
-  model: openai/gpt-oss-120b
-  model-prefix: gptoss
-  runner: gb200
-  precision: fp4
-  framework: dynamo-trt
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      #Right of pareto
-      #P: 1xTP1   D:1xTP4
-      - spec-decoding: "none"
-        conc-list: [ 1, 2, 4, 16, 32, 64, 128 ]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-          - "PREFILL_MAX_NUM_TOKENS=20000"
-          - "PREFILL_MAX_BATCH_SIZE=32"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MAX_NUM_TOKENS=20000"
-          - "DECODE_MAX_BATCH_SIZE=256"
-          - "DECODE_GPU_MEM_FRACTION=0.9"
-
-  # P: 1xTP1   D:4xTP2
-      - spec-decoding: "none"
-        conc-list: [ 16 ]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-          - "PREFILL_MAX_NUM_TOKENS=20000"
-          - "PREFILL_MAX_BATCH_SIZE=32"
-        decode:
-          num-worker: 4
-          tp: 2
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "DECODE_MAX_NUM_TOKENS=20000"
-          - "DECODE_MAX_BATCH_SIZE=32"
-          - "DECODE_GPU_MEM_FRACTION=0.9"
-
-    # P: 1xTP1   D:1xDEP2
-      - spec-decoding: "none"
-        conc-list: [ 256, 512, 1024, 2048, 2560 ]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-          - "PREFILL_MAX_NUM_TOKENS=20000"
-          - "PREFILL_MAX_BATCH_SIZE=32"
-        decode:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MAX_NUM_TOKENS=20000"
-          - "DECODE_MAX_BATCH_SIZE=1536"
-          - "DECODE_GPU_MEM_FRACTION=0.9"
-
-    # P: 1xTP1   D:2xDEP2
-      - spec-decoding: "none"
-        conc-list: [ 512, 1024, 2048, 2560 ]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-          - "PREFILL_MAX_NUM_TOKENS=20000"
-          - "PREFILL_MAX_BATCH_SIZE=32"
-        decode:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MAX_NUM_TOKENS=20000"
-          - "DECODE_MAX_BATCH_SIZE=1536"
-          - "DECODE_GPU_MEM_FRACTION=0.9"
-
-    # P: 1xTP1   D:1xDEP4
-      - spec-decoding: "none"
-        conc-list: [ 256, 1024, 1536 ]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-          - "PREFILL_MAX_NUM_TOKENS=20000"
-          - "PREFILL_MAX_BATCH_SIZE=32"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MAX_NUM_TOKENS=20000"
-          - "DECODE_MAX_BATCH_SIZE=512"
-          - "DECODE_GPU_MEM_FRACTION=0.9"
-
-  # P: 1xTP1   D:3xDEP4
-      - spec-decoding: "none"
-        conc-list: [ 3072 ]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-          - "PREFILL_MAX_NUM_TOKENS=20000"
-          - "PREFILL_MAX_BATCH_SIZE=32"
-        decode:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MAX_NUM_TOKENS=20000"
-          - "DECODE_MAX_BATCH_SIZE=1024"
-          - "DECODE_GPU_MEM_FRACTION=0.9"
-
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # Right side of pareto
-      - spec-decoding: "none"
-        conc-list: [1]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-          - "PREFILL_MAX_NUM_TOKENS=20000"
-          - "PREFILL_MAX_BATCH_SIZE=32"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=2"
-          - "DECODE_MAX_NUM_TOKENS=20000"
-          - "DECODE_MAX_BATCH_SIZE=4"
-          - "DECODE_GPU_MEM_FRACTION=0.9"
-
-      - spec-decoding: "none"
-        conc-list: [2, 4, 8, 16, 32, 64]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-          - "PREFILL_MAX_NUM_TOKENS=20000"
-          - "PREFILL_MAX_BATCH_SIZE=32"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MAX_NUM_TOKENS=20000"
-          - "DECODE_MAX_BATCH_SIZE=128"
-          - "DECODE_GPU_MEM_FRACTION=0.9"
-
-  # Middle of pareto
-  # P: 2xTP1   D:1xTP4
-      - spec-decoding: "none"
-        conc-list: [128, 512]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-          - "PREFILL_MAX_NUM_TOKENS=20000"
-          - "PREFILL_MAX_BATCH_SIZE=32"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MAX_NUM_TOKENS=20000"
-          - "DECODE_MAX_BATCH_SIZE=1024"
-          - "DECODE_GPU_MEM_FRACTION=0.9"
-
-  # P: 2xTP1   D:1xTP2
-      - spec-decoding: "none"
-        conc-list: [256, 384]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-          - "PREFILL_MAX_NUM_TOKENS=20000"
-          - "PREFILL_MAX_BATCH_SIZE=32"
-        decode:
-          num-worker: 1
-          tp: 2
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MAX_NUM_TOKENS=20000"
-          - "DECODE_MAX_BATCH_SIZE=512"
-          - "DECODE_GPU_MEM_FRACTION=0.9"
-
-  # P: 2xTP1   D:1xDEP2
-      - spec-decoding: "none"
-        conc-list: [128, 512]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "PREFILL_NODES=1"
-          - "PREFILL_MAX_NUM_TOKENS=20000"
-          - "PREFILL_MAX_BATCH_SIZE=32"
-        decode:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "DECODE_NODES=1"
-          - "DECODE_MAX_NUM_TOKENS=20000"
-          - "DECODE_MAX_BATCH_SIZE=512"
-          - "DECODE_GPU_MEM_FRACTION=0.9"
-
-# Archive source: nvidia-kimik2.5-8k1k-master.yaml
-# Deprecated Kimi-K2.5/2.6/2.7-Code Single-turn 8k1k entries archived from
-# nvidia-master.yaml. Deprecated after Thursday, August 6, 2026 (see the
-# Deprecation Notice in MODELS.md). Combined with the agentic-coding
-# deprecation of August 3, 2026, this leaves kimik2.5 with no active scenario:
-# the model is fully retired after August 6, 2026.
-# Removed from the active master config so sweep generation no longer selects them.
-
-kimik2.5-int4-b200-vllm:
-  image: vllm/vllm-openai:v0.25.0
-  model: moonshotai/Kimi-K2.5
-  model-prefix: kimik2.5
-  runner: b200
-  precision: int4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-kimik2.5-int4-b300-vllm:
-  image: vllm/vllm-openai:v0.25.0
-  model: moonshotai/Kimi-K2.5
-  model-prefix: kimik2.5
-  runner: b300
-  precision: int4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 64 }
-
-kimik2.5-int4-h200-vllm:
-  image: vllm/vllm-openai:v0.22.0
-  model: moonshotai/Kimi-K2.5
-  model-prefix: kimik2.5
-  runner: h200
-  precision: int4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 4, conc-end: 64 }
-
-kimik2.5-fp4-b200-vllm:
-  image: vllm/vllm-openai:nightly-e2fa28594f7baad142a426b0b6a2cfe2c79201c7
-  model: nvidia/Kimi-K2.5-NVFP4
-  model-prefix: kimik2.5
-  runner: b200
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 1, conc-end: 512 }
-      - { tp: 4, ep: 1, conc-start: 1, conc-end: 512 }
-      - { tp: 8, ep: 8, dp-attn: false, conc-start: 1, conc-end: 512 }
-      - { tp: 4, ep: 4, dp-attn: false, conc-start: 1, conc-end: 512 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-list: [512, 1024] }
-
-# NOTE: At the time of submission, https://docs.vllm.ai/projects/recipes/en/latest/moonshotai/Kimi-K2.5.html
-# does not have a B300-specific recipe, so this config reuses the existing
-# Kimi-K2.5 FP4 B200 vLLM recipe as-is until B300-specific tuning is available.
-kimik2.5-fp4-b300-vllm:
-  image: vllm/vllm-openai:nightly-e2fa28594f7baad142a426b0b6a2cfe2c79201c7
-  model: nvidia/Kimi-K2.5-NVFP4
-  model-prefix: kimik2.5
-  runner: b300
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-list: [1] }
-      - { tp: 4, ep: 1, conc-start: 1, conc-end: 512 }
-      - { tp: 8, ep: 8, dp-attn: false, conc-list: [1] }
-      - { tp: 4, ep: 4, dp-attn: false, conc-start: 1, conc-end: 512 }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 128, conc-end: 512 }
-
-kimik2.5-fp4-b200-dynamo-trt:
-  image: nvcr.io/nvidia/tensorrt-llm/release:1.3.0rc21
-  model: nvidia/Kimi-K2.5-NVFP4
-  model-prefix: kimik2.5
-  runner: cluster:b200-dgxc
-  precision: fp4
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "v1.3.0rc21" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # Non-MTP configurations (STP)
-      - conc-list: [ 4 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/trtllm/kimi-k25-nvfp4/b200-fp4/disagg-B200-1p4d-dep4-tep8-c4-stp.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 24 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/trtllm/kimi-k25-nvfp4/b200-fp4/disagg-B200-1p4d-dep4-tep8-c24-stp.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 5 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/trtllm/kimi-k25-nvfp4/b200-fp4/disagg-B200-1p5d-dep4-tep4-c5-stp.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 30 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/trtllm/kimi-k25-nvfp4/b200-fp4/disagg-B200-1p5d-dep4-tep4-c30-stp.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 60 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/trtllm/kimi-k25-nvfp4/b200-fp4/disagg-B200-1p5d-dep4-tep4-c60-stp.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 115 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/trtllm/kimi-k25-nvfp4/b200-fp4/disagg-B200-1p5d-dep4-tep4-c115-stp.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 195 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/trtllm/kimi-k25-nvfp4/b200-fp4/disagg-B200-1p5d-dep4-tep4-c195-stp.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 308 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/trtllm/kimi-k25-nvfp4/b200-fp4/disagg-B200-1p1d-dep4-dep8-c308-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [ 615 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/trtllm/kimi-k25-nvfp4/b200-fp4/disagg-B200-2p1d-dep4-dep8-c615-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [ 1127 ]
-        prefill:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/trtllm/kimi-k25-nvfp4/b200-fp4/disagg-B200-3p1d-dep4-dep8-c1127-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [ 2151 ]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/trtllm/kimi-k25-nvfp4/b200-fp4/disagg-B200-4p1d-dep4-dep8-c2151-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-kimik2.5-fp4-gb200-dynamo-trt:
-  image: nvcr.io#nvidia/ai-dynamo/tensorrtllm-runtime:1.3.0-dev.1-cuda13
-  model: nvidia/Kimi-K2.5-NVFP4
-  model-prefix: kimik2.5
-  runner: gb200
-  precision: fp4
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "v1.3.0-dev.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # Non-MTP configurations (default spec_decoding="none")
-      - conc-list: [ 8 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen4tep8_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen4tep8_batch1_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 24 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen4tep8_batch4_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen4tep8_batch4_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 5 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch1_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 30 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch4_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch4_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 60 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch8_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch8_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 105 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch16_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch16_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 333 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx2dep4_gen1dep32_batch8_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx2dep4_gen1dep32_batch8_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 666 ]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx4dep4_gen1dep32_batch16_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx4dep4_gen1dep32_batch16_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 1229 ]
-        prefill:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx4dep4_gen1dep16_batch64_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx4dep4_gen1dep16_batch64_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [ 1229 ]
-        prefill:
-          num-worker: 6
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx6dep4_gen1dep32_batch32_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx6dep4_gen1dep32_batch32_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 2253 ]
-        prefill:
-          num-worker: 7
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx7dep4_gen1dep16_batch128_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx7dep4_gen1dep16_batch128_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [ 4301 ]
-        prefill:
-          num-worker: 9
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx9dep4_gen1dep16_batch256_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb200Nvfp4/ISL8K_OSL1K/STP/ctx9dep4_gen1dep16_batch256_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-kimik2.5-fp4-gb300-dynamo-trt:
-  image: nvcr.io#nvidia/ai-dynamo/tensorrtllm-runtime:1.3.0-dev.1-cuda13
-  model: nvidia/Kimi-K2.5-NVFP4
-  model-prefix: kimik2.5
-  runner: gb300
-  precision: fp4
-  framework: dynamo-trt
-  router: { name: dynamo-router, version: "v1.3.0-dev.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # Non-MTP configurations (default spec_decoding="none")
-      - conc-list: [ 4 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen4tep8_batch1_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen4tep8_batch1_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 12 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen4tep8_batch2_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen4tep8_batch2_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 24 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen4tep8_batch4_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen4tep8_batch4_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [ 30 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch4_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch4_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 60 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch8_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch8_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 115 ]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch16_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx1dep4_gen5tep4_batch16_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [ 333 ]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx2dep4_gen1dep32_batch8_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx2dep4_gen1dep32_batch8_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 615 ]
-        prefill:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx3dep4_gen1dep32_batch16_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx3dep4_gen1dep32_batch16_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 1229 ]
-        prefill:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx5dep4_gen1dep32_batch32_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx5dep4_gen1dep32_batch32_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-      - conc-list: [ 2253 ]
-        prefill:
-          num-worker: 6
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx6dep4_gen1dep16_batch128_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx6dep4_gen1dep16_batch128_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-      - conc-list: [ 2151 ]
-        prefill:
-          num-worker: 8
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          # https://github.com/NVIDIA/srt-slurm/blob/sa-submission-q2-2026/recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx8dep4_gen1dep32_batch64_eplb0_mtp0.yaml
-          - "CONFIG_FILE=recipes/kimi2.5/trtllm_dynamo/disagg/gb300Nvfp4/ISL8K_OSL1K/STP/ctx8dep4_gen1dep32_batch64_eplb0_mtp0.yaml"
-        decode:
-          num-worker: 1
-          tp: 32
-          ep: 32
-          dp-attn: true
-
-kimik2.5-fp4-gb200-dynamo-vllm:
-  image: vllm/vllm-openai:v0.21.0
-  model: nvidia/Kimi-K2.5-NVFP4
-  model-prefix: kimik2.5
-  runner: gb200
-  precision: fp4
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.2.1" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [128]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.5/vllm/gb200-fp4/8k1k/disagg-1p4d-dep4-tep4-stp.yaml"
-        decode:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [4, 8, 16, 32, 256]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.5/vllm/gb200-fp4/8k1k/disagg-1p4d-dep4-tp8-stp.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [1024]
-        prefill:
-          num-worker: 3
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.5/vllm/gb200-fp4/8k1k/disagg-3p1d-dep4-dep16-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 1
-          ep: 16
-          dp-attn: true
-      - conc-list: [3072]
-        prefill:
-          num-worker: 6
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.5/vllm/gb200-fp4/8k1k/disagg-6p1d-dep4-dep16-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 1
-          ep: 16
-          dp-attn: true
-      - conc-list: [6144]
-        prefill:
-          num-worker: 8
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.5/vllm/gb200-fp4/8k1k/disagg-8p1d-dep4-dep16-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 1
-          ep: 16
-          dp-attn: true
-
-kimik2.5-fp4-gb300-dynamo-vllm:
-  image: vllm/vllm-openai:v0.21.0
-  model: nvidia/Kimi-K2.5-NVFP4
-  model-prefix: kimik2.5
-  runner: gb300-nv
-  precision: fp4
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.3.0.dev20260601" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [4]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.5/vllm/gb300-fp4/8k1k/disagg-1p4d-dep4-tp8-stp.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [32, 128]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.5/vllm/gb300-fp4/8k1k/disagg-1p8d-dep4-tp4-stp.yaml"
-        decode:
-          num-worker: 8
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [2048]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.5/vllm/gb300-fp4/8k1k/disagg-2p1d-dep4-dep24-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 1
-          ep: 24
-          dp-attn: true
-      - conc-list: [2048]
-        prefill:
-          num-worker: 3
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.5/vllm/gb300-fp4/8k1k/disagg-3p1d-dep4-dep16-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 1
-          ep: 16
-          dp-attn: true
-      - conc-list: [3072]
-        prefill:
-          num-worker: 4
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.5/vllm/gb300-fp4/8k1k/disagg-4p1d-dep4-dep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 1
-          ep: 8
-          dp-attn: true
-      - conc-list: [15360]
-        prefill:
-          num-worker: 8
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.5/vllm/gb300-fp4/8k1k/disagg-8p1d-dep4-dep24-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 1
-          ep: 24
-          dp-attn: true
-
-# Archive source: nvidia-kimik2.6-8k1k-master.yaml
-# Deprecated Kimi-K2.6 Single-turn 8k1k entry archived from nvidia-master.yaml.
-# The August 6, 2026 retirement covered Kimi-K2.5/2.6/2.7-Code, but the
-# separate kimik2.6 prefix was missed. Archived on September 17, 2026.
-# Keep the referenced srt-slurm recipes and launcher support for historical use.
-
-kimik2.6-fp4-b200-dynamo-vllm:
-  image: vllm/vllm-openai:v0.25.1
-  model: nvidia/Kimi-K2.6-NVFP4
-  model-prefix: kimik2.6
-  runner: cluster:b200-nscale
-  precision: fp4
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.3.0.dev20260721" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [1]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.6/vllm/b200-fp4/8k1k/disagg-1p1d-dep8-tp8-c1-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [32]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.6/vllm/b200-fp4/8k1k/disagg-1p8d-dep4-tp4-c32-stp.yaml"
-        decode:
-          num-worker: 8
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [128]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.6/vllm/b200-fp4/8k1k/disagg-1p8d-dep4-tp4-c128-stp.yaml"
-        decode:
-          num-worker: 8
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [512]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.6/vllm/b200-fp4/8k1k/disagg-1p4d-dep4-tp4-c512-stp.yaml"
-        decode:
-          num-worker: 4
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [2048]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.6/vllm/b200-fp4/8k1k/disagg-1p1d-dep8-dep8-c2048-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 1
-          ep: 8
-          dp-attn: true
-      - conc-list: [8192]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 8
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/kimik2.6/vllm/b200-fp4/8k1k/disagg-2p1d-dep8-dep8-c8192-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 1
-          ep: 8
-          dp-attn: true
-
-# Archive source: nvidia-minimaxm2.5-m2.7-master.yaml
-# Deprecated MiniMax M2.5/M2.7 entries archived from nvidia-master.yaml.
-# Removed from the active master config so sweep generation no longer selects them.
-
-
-minimaxm2.5-fp8-b200-vllm:
-  image: vllm/vllm-openai:v0.22.0
-  model: MiniMaxAI/MiniMax-M2.5
-  model-prefix: minimaxm2.5
-  runner: b200
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 2, ep: 2, conc-start: 512, conc-end: 512 }
-      - { tp: 4, conc-start: 4, conc-end: 128 }
-      - { tp: 4, ep: 4, conc-start: 256, conc-end: 512 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 2, conc-start: 4, conc-end: 512 }
-      - { tp: 4, conc-start: 4, conc-end: 512 }
-
-  # NOTE: At the time of submission, https://docs.vllm.ai/projects/recipes/en/latest/MiniMax/MiniMax-M2.html
-  # does not have a B300-specific recipe, so this config reuses the existing
-  # MiniMax-M2.5 FP8 B200 vLLM recipe as-is until B300-specific tuning is available.
-
-# Diverged from minimaxm2.5-fp8-b200-vllm (agentic-coding sibling). Reasons below;
-# the original minimaxm2.5-fp8-b200-vllm entry is left identical to origin/main so
-# its fixed-seq-len sweep is unaffected.
-#   - image: 'vllm/vllm-openai:v0.19.0-cu130' -> 'vllm/vllm-openai:v0.19.1'
-#   - runner: 'b200' -> 'b200-dgxc'
-minimaxm2.5-fp8-b300-vllm:
-  image: vllm/vllm-openai:v0.21.0
-  model: MiniMaxAI/MiniMax-M2.5
-  model-prefix: minimaxm2.5
-  runner: b300
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 4, conc-end: 128 }
-      - { tp: 4, ep: 4, conc-start: 256, conc-end: 512 }
-      - { tp: 2, ep: 2, conc-start: 512, conc-end: 1024 }
-      - { tp: 2, ep: 2, dp-attn: true, conc-start: 1024, conc-end: 1024 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 4, conc-end: 16 }
-      - { tp: 2, conc-start: 64, conc-end: 256 }
-      - { tp: 4, conc-start: 4, conc-end: 8 }
-
-# Diverged from minimaxm2.5-fp8-b300-vllm (agentic-coding sibling). Reasons below;
-# the original minimaxm2.5-fp8-b300-vllm entry is left identical to origin/main so
-# its fixed-seq-len sweep is unaffected.
-#   - image: 'vllm/vllm-openai:v0.19.0-cu130' -> 'vllm/vllm-openai:v0.19.1'
-minimaxm2.5-fp4-b200-vllm:
-  image: vllm/vllm-openai:v0.22.0
-  model: nvidia/MiniMax-M2.5-NVFP4
-  model-prefix: minimaxm2.5
-  runner: b200
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 4, conc-end: 16 }
-      - { tp: 2, conc-start: 16, conc-end: 16 }
-      - { tp: 2, ep: 2, conc-start: 128, conc-end: 128 }
-      - { tp: 2, ep: 2, dp-attn: true, conc-start: 256, conc-end: 1024 }
-      - { tp: 4, conc-start: 4, conc-end: 16 }
-      - { tp: 4, ep: 4, conc-start: 64, conc-end: 128 }
-      - { tp: 8, conc-start: 4, conc-end: 8 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 4, conc-end: 32 }
-      - { tp: 1, conc-start: 256, conc-end: 256 }
-      - { tp: 2, ep: 2, conc-start: 128, conc-end: 512 }
-      - { tp: 4, conc-start: 4, conc-end: 8 }
-      - { tp: 8, conc-start: 4, conc-end: 4 }
-
-  # NOTE: At the time of submission, https://docs.vllm.ai/projects/recipes/en/latest/MiniMax/MiniMax-M2.html
-  # does not have a B300-specific recipe, so this config reuses the existing
-  # MiniMax-M2.5 FP4 B200 vLLM recipe as-is until B300-specific tuning is available.
-
-# Diverged from minimaxm2.5-fp4-b200-vllm (agentic-coding sibling). Metadata is
-# identical to origin/main's minimaxm2.5-fp4-b200-vllm; the split exists because this
-# PR adds an agentic-coding scenarios block that differs from main
-# (either main had none or had a different conc/offload sweep).
-# The original minimaxm2.5-fp4-b200-vllm entry stays byte-identical to origin/main.
-minimaxm2.5-fp4-b200-trt:
-  image: nvcr.io#nvidia/tensorrt-llm/release:1.3.0rc18
-  model: nvidia/MiniMax-M2.5-NVFP4
-  model-prefix: minimaxm2.5
-  runner: b200
-  precision: fp4
-  framework: trt
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 2, ep: 2, dp-attn: true, conc-start: 1024, conc-end: 1024 }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 1024, conc-end: 1024 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 128, conc-end: 1024 }
-      - { tp: 1, ep: 1, conc-start: 4, conc-end: 8 }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 8 }
-      - { tp: 8, ep: 2, conc-start: 4, conc-end: 4 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 256, conc-end: 1024 }
-      - { tp: 2, ep: 2, dp-attn: true, conc-start: 256, conc-end: 256 }
-      - { tp: 1, ep: 1, conc-start: 4, conc-end: 32 }
-      - { tp: 4, ep: 2, conc-start: 4, conc-end: 4 }
-      - { tp: 8, ep: 2, conc-start: 4, conc-end: 4 }
-
-minimaxm2.5-fp4-b300-vllm:
-  image: vllm/vllm-openai:v0.21.0
-  model: nvidia/MiniMax-M2.5-NVFP4
-  model-prefix: minimaxm2.5
-  runner: b300
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 4, conc-end: 8 }
-      - { tp: 2, ep: 2, conc-start: 128, conc-end: 128 }
-      - { tp: 2, ep: 2, dp-attn: true, conc-start: 256, conc-end: 2048 }
-      - { tp: 4, conc-start: 8, conc-end: 8 }
-      - { tp: 4, ep: 4, conc-start: 64, conc-end: 128 }
-      - { tp: 8, conc-start: 4, conc-end: 8 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 1, conc-start: 4, conc-end: 256 }
-      - { tp: 2, ep: 2, dp-attn: true, conc-start: 512, conc-end: 512 }
-      - { tp: 4, conc-start: 4, conc-end: 8 }
-      - { tp: 8, conc-start: 4, conc-end: 4 }
-
-minimaxm2.5-fp4-b300-trt:
-  image: nvcr.io#nvidia/tensorrt-llm/release:1.3.0rc18
-  model: nvidia/MiniMax-M2.5-NVFP4
-  model-prefix: minimaxm2.5
-  runner: b300
-  precision: fp4
-  framework: trt
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 2, ep: 2, dp-attn: true, conc-start: 2048, conc-end: 2048 }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 1024, conc-end: 2048 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 128, conc-end: 1024 }
-      - { tp: 1, ep: 1, dp-attn: false, conc-start: 4, conc-end: 8 }
-      - { tp: 4, ep: 2, dp-attn: false, conc-start: 4, conc-end: 8 }
-      - { tp: 8, ep: 2, dp-attn: false, conc-start: 4, conc-end: 4 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 2, ep: 2, dp-attn: true, conc-start: 128, conc-end: 512 }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 128, conc-end: 128 }
-      - { tp: 1, ep: 1, dp-attn: false, conc-start: 4, conc-end: 16 }
-      - { tp: 4, ep: 2, dp-attn: false, conc-start: 4, conc-end: 4 }
-      - { tp: 8, ep: 2, dp-attn: false, conc-start: 4, conc-end: 4 }
-
-minimaxm2.5-fp8-h100-vllm:
-  image: vllm/vllm-openai:v0.22.0
-  model: MiniMaxAI/MiniMax-M2.5
-  model-prefix: minimaxm2.5
-  runner: h100
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 8, conc-start: 4, conc-end: 128 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 8, conc-start: 4, conc-end: 128 }
-
-# Diverged from minimaxm2.5-fp8-h100-vllm (agentic-coding sibling). Metadata is
-# identical to origin/main's minimaxm2.5-fp8-h100-vllm; the split exists because this
-# PR adds an agentic-coding scenarios block that differs from main
-# (either main had none or had a different conc/offload sweep).
-# The original minimaxm2.5-fp8-h100-vllm entry stays byte-identical to origin/main.
-minimaxm2.5-fp8-h200-vllm:
-  image: vllm/vllm-openai:v0.22.0
-  model: MiniMaxAI/MiniMax-M2.5
-  model-prefix: minimaxm2.5
-  runner: h200
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 256 }
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 256 }
-
-# Diverged from minimaxm2.5-fp8-h200-vllm (agentic-coding sibling). Metadata is
-# identical to origin/main's minimaxm2.5-fp8-h200-vllm; the split exists because this
-# PR adds an agentic-coding scenarios block that differs from main
-# (either main had none or had a different conc/offload sweep).
-# The original minimaxm2.5-fp8-h200-vllm entry stays byte-identical to origin/main.
-minimaxm2.5-fp4-gb300-dynamo-vllm:
-  image: vllm/vllm-openai:v0.20.1
-  model: nvidia/MiniMax-M2.5-NVFP4
-  model-prefix: minimaxm2.5
-  runner: gb300-nv
-  precision: fp4
-  framework: dynamo-vllm
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - conc-list: [2, 4, 16]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4-1p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [4, 8, 16, 64]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4-1p2d.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [32, 64, 128]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4ep-1p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [64, 128, 256, 512, 1024]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4ep-1p3d.yaml"
-        decode:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [2048]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/dep2-2p3d.yaml"
-        decode:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-      - conc-list: [6144, 8192]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/dep2-2p3d-c6144.yaml"
-        decode:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-      - conc-list: [1024, 2048, 4096]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/dep8-2p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [2, 4, 8, 16]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/tp4-1p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [32, 64, 128, 256]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/tp4ep-1p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [64, 128]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/tp4ep-2p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [256]
-        prefill:
-          num-worker: 4
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/dep4-4p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-      - conc-list: [1024, 2048]
-        prefill:
-          num-worker: 4
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/dep8-4p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-minimaxm2.5-fp8-gb200-dynamo-vllm:
-  image: vllm/vllm-openai:v0.20.1
-  model: MiniMaxAI/MiniMax-M2.5
-  model-prefix: minimaxm2.5
-  runner: gb200
-  precision: fp8
-  framework: dynamo-vllm
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - conc-list: [1, 4, 8, 16, 32, 64]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p1d-tp4.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [2, 32, 64, 128, 256, 512]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p2d-tp4.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [1024]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p3d-tp4ep.yaml"
-        decode:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [512, 1024]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-2p1d-dep8.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [4096]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p4d-dep2.yaml"
-        decode:
-          num-worker: 4
-          tp: 2
-          ep: 2
-          dp-attn: true
-      - conc-list: [4096, 8192]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-2p3d-dep4.yaml"
-        decode:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [1, 4, 8, 16, 32, 64, 128]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/8k1k/disagg-gb200-1p1d-tp4.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [256, 512]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/8k1k/disagg-gb200-1p1d-tp4ep.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [1024, 2048, 4096]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/8k1k/disagg-gb200-3p2d-dep4.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-
-
-minimaxm2.5-fp8-b200-dynamo-vllm:
-  image: vllm/vllm-openai:v0.20.1
-  model: MiniMaxAI/MiniMax-M2.5
-  model-prefix: minimaxm2.5
-  runner: cluster:b200-dgxc
-  precision: fp8
-  framework: dynamo-vllm
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - conc-list: [32, 64]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp8/1k1k/disagg-b200-1p1d-tp4ep.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [128]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp8/1k1k/tp4ep.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [1024]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp8/1k1k/disagg-b200-1p3d-tp4ep.yaml"
-        decode:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [512, 1024]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp8/1k1k/disagg-b200-2p1d-dep8.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [512]
-        prefill:
-          num-worker: 4
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp8/1k1k/dep8.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [4096]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp8/1k1k/disagg-b200-1p4d-dep2-hi-conc.yaml"
-        decode:
-          num-worker: 4
-          tp: 2
-          ep: 2
-          dp-attn: true
-      - conc-list: [4096, 8192]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp8/1k1k/disagg-b200-2p3d-dep4-hi-conc.yaml"
-        decode:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [16, 32, 64, 128]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp8/8k1k/disagg-b200-1p1d-tp4ep.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [256, 512]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp8/8k1k/disagg-b200-1p1d-tp4ep-hi-conc.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [1024, 2048]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp8/8k1k/disagg-b200-3p2d-dep4.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-
-minimaxm2.5-fp4-gb200-dynamo-vllm:
-  image: vllm/vllm-openai:v0.20.1
-  model: nvidia/MiniMax-M2.5-NVFP4
-  model-prefix: minimaxm2.5
-  runner: gb200
-  precision: fp4
-  framework: dynamo-vllm
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - conc-list: [16]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200/1k1k/tp4-1p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [4, 8, 16, 32, 64]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200/1k1k/tp4-1p2d.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [32, 64, 128]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200/1k1k/tp4ep-1p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [256]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200/1k1k/tp4ep-1p1d-hi-conc.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [128, 256]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200/1k1k/tp4ep-1p2d.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [64, 128, 256, 512]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200/1k1k/tp4ep-1p3d.yaml"
-        decode:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [1024, 2048]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200/1k1k/tp4ep-1p3d-hi-conc.yaml"
-        decode:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [256, 1024]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200/1k1k/tp4ep-2p3d.yaml"
-        decode:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [2048, 4096, 8192]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200/1k1k/dep2-1p2d.yaml"
-        decode:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-      - conc-list: [4096]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200/1k1k/dep2-2p3d.yaml"
-        decode:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-      - conc-list: [6144]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200/1k1k/dep2-2p3d-c6144.yaml"
-        decode:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-      - conc-list: [1024, 2048]
-        prefill:
-          num-worker: 3
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200/1k1k/dep4-3p2d.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-      - conc-list: [1024, 2048, 4096]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200/1k1k/dep8-2p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [4, 8, 16]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200/8k1k/tp4-1p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [32, 64]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200/8k1k/tp4ep-1p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [256, 512, 1024]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200/8k1k/tp4ep-1p1d-hi-conc.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [256, 512, 1024, 2048]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200/8k1k/dep4-2p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-
-minimaxm2.5-fp4-b200-dynamo-vllm:
-  image: vllm/vllm-openai:v0.20.1
-  model: nvidia/MiniMax-M2.5-NVFP4
-  model-prefix: minimaxm2.5
-  runner: cluster:b200-dgxc
-  precision: fp4
-  framework: dynamo-vllm
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - conc-list: [16]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp4/1k1k/tp4-1p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [4, 8, 16, 32, 64]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp4/1k1k/tp4-1p2d.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [32, 64, 128]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp4/1k1k/tp4ep-1p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [256]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp4/1k1k/tp4ep-1p1d-hi-conc.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [128, 256]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp4/1k1k/tp4ep-1p2d.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [64, 128, 256, 512]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp4/1k1k/tp4ep-1p3d.yaml"
-        decode:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [1024, 2048]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp4/1k1k/tp4ep-1p3d-hi-conc.yaml"
-        decode:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [256, 1024]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp4/1k1k/tp4ep-2p3d.yaml"
-        decode:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [2048, 4096, 8192]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp4/1k1k/dep2-1p2d.yaml"
-        decode:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-      - conc-list: [4096]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp4/1k1k/dep2-2p3d.yaml"
-        decode:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-      - conc-list: [6144]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp4/1k1k/dep2-2p3d-c6144.yaml"
-        decode:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-      - conc-list: [1024, 2048]
-        prefill:
-          num-worker: 3
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp4/1k1k/dep4-3p2d.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-      - conc-list: [1024, 2048, 4096]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp4/1k1k/dep8-2p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [4, 8, 16]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp4/8k1k/tp4-1p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [32, 64]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp4/8k1k/tp4ep-1p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [256, 512, 1024]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp4/8k1k/tp4ep-1p1d-hi-conc.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [256, 512, 1024, 2048]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-b200-fp4/8k1k/dep4-2p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-
-minimaxm2.5-fp4-b300-dynamo-vllm:
-  image: vllm/vllm-openai:v0.20.1
-  model: nvidia/MiniMax-M2.5-NVFP4
-  model-prefix: minimaxm2.5
-  runner: b300
-  precision: fp4
-  framework: dynamo-vllm
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - conc-list: [4, 16]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4-1p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [4]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp8-1p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [8, 16]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4-1p2d.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [32, 64, 128]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4ep-1p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [64, 128, 256, 1024]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4ep-1p3d.yaml"
-        decode:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [4096]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/dep2-1p2d.yaml"
-        decode:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-      - conc-list: [2048, 4096]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/dep2-2p3d.yaml"
-        decode:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-      - conc-list: [6144, 8192]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/dep2-2p3d-c6144.yaml"
-        decode:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-      - conc-list: [1024, 1536, 2048, 4096]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/dep8-2p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [2, 4, 8, 16]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/tp4-1p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [4]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/tp8-1p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [32, 128]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/tp4ep-1p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [64, 128, 256, 512]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/tp4ep-2p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [384, 512]
-        prefill:
-          num-worker: 4
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/dep4-4p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-      - conc-list: [384]
-        prefill:
-          num-worker: 4
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/dep8-4p1d.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-minimaxm2.5-fp8-b300-dynamo-vllm:
-  image: vllm/vllm-openai:v0.20.1
-  model: MiniMaxAI/MiniMax-M2.5
-  model-prefix: minimaxm2.5
-  runner: b300
-  precision: fp8
-  framework: dynamo-vllm
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - conc-list: [8, 16, 32, 64, 128]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-b300-1p1d-tp4.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [32, 64, 128, 256, 512]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-b300-1p2d-tp4.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [256, 512, 1024]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-b300-1p2d-tp4ep.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [256, 512, 1024]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-b300-2p1d-dep8.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [512, 1024, 2048]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-b300-2p2d-dep4.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-      - conc-list: [4096, 8192]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-b300-2p2d-dep4-hi-conc.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-      - conc-list: [1024]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-b300-2p3d-dep2.yaml"
-        decode:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [16, 64, 128]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-b300-1p1d-tp4ep.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [256, 512]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-b300-1p1d-tp4ep-hi-conc.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [32]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-b300-2p1d-tp2.yaml"
-        decode:
-          num-worker: 1
-          tp: 2
-          ep: 1
-          dp-attn: false
-      - conc-list: [64, 128, 256, 512]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-b300-2p1d-tp4ep.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [64]
-        prefill:
-          num-worker: 3
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-b300-3p1d-tp4.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [256, 512]
-        prefill:
-          num-worker: 3
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-b300-3p1d-dep4.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-      - conc-list: [1024, 2048]
-        prefill:
-          num-worker: 3
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-b300-3p1d-dep4-hi-conc.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-      - conc-list: [512, 1024, 2048]
-        prefill:
-          num-worker: 5
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-b300-5p2d-dep4.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-
-minimaxm2.5-fp8-gb300-dynamo-vllm:
-  image: vllm/vllm-openai:v0.20.1
-  model: MiniMaxAI/MiniMax-M2.5
-  model-prefix: minimaxm2.5
-  runner: gb300-nv
-  precision: fp8
-  framework: dynamo-vllm
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 1024
-      osl: 1024
-      search-space:
-      - conc-list: [8, 16, 32, 64, 128]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-1p1d-tp4.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [32, 64, 128, 256, 512]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-1p2d-tp4.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [256, 512, 1024]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-1p2d-tp4ep.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [256, 512, 1024]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-2p1d-dep8.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [512, 1024, 2048]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-2p2d-dep4.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-      - conc-list: [4096, 8192]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-2p2d-dep4-hi-conc.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-      - conc-list: [1024]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-2p3d-dep2.yaml"
-        decode:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [16, 64, 128]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-1p1d-tp4ep.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [256, 512]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-1p1d-tp4ep-hi-conc.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [32]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-2p1d-tp2.yaml"
-        decode:
-          num-worker: 1
-          tp: 2
-          ep: 1
-          dp-attn: false
-      - conc-list: [64, 128, 256, 512]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-2p1d-tp4ep.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [64]
-        prefill:
-          num-worker: 3
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-3p1d-tp4.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [256, 512]
-        prefill:
-          num-worker: 3
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-3p1d-dep4.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-      - conc-list: [1024, 2048]
-        prefill:
-          num-worker: 3
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-3p1d-dep4-hi-conc.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-      - conc-list: [512, 1024, 2048]
-        prefill:
-          num-worker: 5
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-5p2d-dep4.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-
-# Archive source: nvidia-minimaxm3-8k1k-master.yaml
-# Deprecated MiniMax-M3 Single-turn 8k1k entries archived from nvidia-master.yaml.
-# The scenario was deprecated after Monday, August 3, 2026 (see the Deprecation
-# Notice in MODELS.md). Agentic coding remains active for this model.
-# Removed from the active master config so sweep generation no longer selects them.
-
-
-minimaxm3-fp8-h100-vllm:
-  image: vllm/vllm-openai:minimax-m3
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: h100
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 64 }
-      - { tp: 8, ep: 8, conc-start: 1, conc-end: 256 }
-
-minimaxm3-fp8-h200-vllm:
-  image: vllm/vllm-openai:minimax-m3
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: h200
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 32 }
-      - { tp: 4, ep: 4, conc-start: 1, conc-end: 256 }
-      - { tp: 8, conc-start: 1, conc-end: 128 }
-      - { tp: 8, ep: 8, conc-start: 1, conc-end: 256 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 512 }
-
-minimaxm3-fp4-b200-dynamo-vllm:
-  image: vllm/vllm-openai:nightly-5e35a6f4f9bbc217c599692157ca985c894373f7
-  model: nvidia/MiniMax-M3-NVFP4
-  model-prefix: minimaxm3
-  runner: cluster:b200-dgxc
-  precision: fp4
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.3.0.dev20260710" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [1, 4, 8, 16]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b200-fp4/8k1k/disagg-1p1d-dep2-tp4-c1-c16-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [64]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b200-fp4/8k1k/disagg-1p2d-dep2-tp4-c64-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [128]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b200-fp4/8k1k/disagg-2p2d-dep2-tp4-c128-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [256]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b200-fp4/8k1k/disagg-3p2d-dep2-tp4-c256-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [512]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b200-fp4/8k1k/disagg-3p2d-dep2-tp4-c512-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [1024]
-        prefill:
-          num-worker: 4
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b200-fp4/8k1k/disagg-4p2d-dep2-tp4-c1024-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-# MiniMax-M3 NVFP4 disagg sweep on the same B300 topology matrix as the MXFP8
-# baseline above. The image includes vLLM PR #46380, so no runtime patch is
-# needed.
-minimaxm3-fp8-b300-dynamo-vllm:
-  image: vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: b300
-  precision: fp8
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.3.0.dev20260614" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [256, 512]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp8/8k1k/disagg-2p2d-dep2-dep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [16]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp8/8k1k/disagg-2p2d-dep2-tep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: false
-      - conc-list: [4096]
-        prefill:
-          num-worker: 4
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp8/8k1k/disagg-4p2d-dep2-dep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [1, 4, 8, 16]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp8/8k1k/disagg-1p1d-dep2-tp4-marlin-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [1024, 2048]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp8/8k1k/disagg-3p2d-dep2-dep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [16, 32, 64, 128]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp8/8k1k/disagg-1p4d-dep2-tep4-stp.yaml"
-        decode:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [16]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp8/8k1k/disagg-1p2d-dep2-tep4-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: false
-      - conc-list: [4]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp8/8k1k/disagg-1p4d-dep2-tep8-stp.yaml"
-        decode:
-          num-worker: 4
-          tp: 8
-          ep: 8
-          dp-attn: false
-
-# MiniMax-M3 NVFP4 disagg sweep on the same B300 topology matrix as the MXFP8
-# baseline above. The image includes vLLM PR #46380, so no runtime patch is
-# needed.
-minimaxm3-fp4-b300-dynamo-vllm-8k1k-tp1:
-  image: vllm/vllm-openai:nightly-5e35a6f4f9bbc217c599692157ca985c894373f7
-  model: nvidia/MiniMax-M3-NVFP4
-  model-prefix: minimaxm3
-  runner: b300
-  precision: fp4
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.3.0.dev20260713" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [8, 16, 24, 32, 48, 64]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/8k1k/disagg-1p1d-tp1-tp4-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [1, 2, 4, 8, 16]
-        prefill:
-          num-worker: 1
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/8k1k/disagg-1p1d-tp1-tp8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - conc-list: [128]
-        prefill:
-          num-worker: 2
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/8k1k/disagg-2p1d-tp1-tp4-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [256]
-        prefill:
-          num-worker: 3
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/8k1k/disagg-3p1d-tp1-tp4-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - conc-list: [512]
-        prefill:
-          num-worker: 5
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/8k1k/disagg-5p2d-tp1-dep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - conc-list: [768, 1024]
-        prefill:
-          num-worker: 8
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/8k1k/disagg-8p2d-tp1-dep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-# Compare the refreshed image on a one-node max-throughput topology: four
-# TP1 prefills colocated with one TEP4 decode worker.
-minimaxm3-fp4-b300-dynamo-vllm-8k1k-tp1-tep4:
-  image: vllm/vllm-openai:nightly-5e35a6f4f9bbc217c599692157ca985c894373f7
-  model: nvidia/MiniMax-M3-NVFP4
-  model-prefix: minimaxm3
-  runner: b300
-  precision: fp4
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.3.0.dev20260713" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [4096]
-        prefill:
-          num-worker: 4
-          tp: 1
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/8k1k/disagg-4p1d-tp1-tep4-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: false
-
-# Preserve the previously proven max-throughput point exactly, while keeping
-# it separate from the legacy mixed 1k1k/8k1k entry so 1k1k is not rerun.
-minimaxm3-fp4-b300-dynamo-vllm-8k1k-legacy-max-tput:
-  image: vllm/vllm-openai:vllm-minimax-m3-perf-x86_64-13.0.1-8b00f41
-  model: nvidia/MiniMax-M3-NVFP4
-  model-prefix: minimaxm3
-  runner: b300
-  precision: fp4
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.3.0.dev20260614" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - conc-list: [4096]
-        prefill:
-          num-worker: 4
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/8k1k/disagg-4p2d-dep2-tep4-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: false
-
-minimaxm3-fp4-b200-dynamo-vllm-mtp:
-  image: vllm/vllm-openai:nightly-5e35a6f4f9bbc217c599692157ca985c894373f7
-  model: nvidia/MiniMax-M3-NVFP4
-  model-prefix: minimaxm3
-  runner: cluster:b200-dgxc
-  precision: fp4
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.3.0.dev20260710" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - spec-decoding: "mtp"
-        conc-list: [12, 24, 48]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b200-fp4/8k1k/disagg-1p6d-tp4-tp4-eagle3-mtp.yaml"
-        decode:
-          num-worker: 6
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [4, 8, 16, 32, 64]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b200-fp4/8k1k/disagg-1p4d-tp4-tp4-eagle3-mtp.yaml"
-        decode:
-          num-worker: 4
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [32, 64, 96]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b200-fp4/8k1k/disagg-1p2d-tp4-tp4-eagle3-mtp.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [192, 256, 512]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b200-fp4/8k1k/disagg-1p1d-tp4-tp4-eagle3-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-      - spec-decoding: "mtp"
-        conc-list: [12, 24, 48]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b200-fp4/8k1k/disagg-1p3d-tp4-tp4-eagle3-mtp.yaml"
-        decode:
-          num-worker: 3
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-
-      - spec-decoding: "mtp"
-        conc-list: [256, 512]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b200-fp4/8k1k/disagg-1p1d-tp4-dep4-eagle3-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-
-minimaxm3-fp4-b300-dynamo-vllm-mtp:
-  image: vllm/vllm-openai:nightly-5e35a6f4f9bbc217c599692157ca985c894373f7
-  model: nvidia/MiniMax-M3-NVFP4
-  model-prefix: minimaxm3
-  runner: b300
-  precision: fp4
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.3.0.dev20260710" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - spec-decoding: "mtp"
-        conc-list: [24, 48]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/8k1k/disagg-1p6d-dep2-tp4-eagle3-mtp.yaml"
-        decode:
-          num-worker: 6
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [4, 8, 64]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/8k1k/disagg-1p4d-dep2-tp4-eagle3-mtp.yaml"
-        decode:
-          num-worker: 4
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [4, 32, 64]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/8k1k/disagg-1p2d-dep2-tp4-eagle3-mtp.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [64]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/8k1k/disagg-1p1d-dep2-tp4-eagle3-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-      - spec-decoding: "mtp"
-        conc-list: [192]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/8k1k/disagg-2p3d-dep2-tp4-eagle3-mtp.yaml"
-        decode:
-          num-worker: 3
-          tp: 4
-          ep: 1
-          dp-attn: false
-
-minimaxm3-fp4-b300-dynamo-vllm-mtp-legacy-dep4:
-  image: vllm/vllm-openai:nightly-5e35a6f4f9bbc217c599692157ca985c894373f7
-  model: nvidia/MiniMax-M3-NVFP4
-  model-prefix: minimaxm3
-  runner: b300
-  precision: fp4
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.3.0.dev20260710" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - spec-decoding: "mtp"
-        conc-list: [256, 512]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/b300-fp4/8k1k/disagg-2p1d-dep2-dep4-eagle3-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-
-# MiniMax-M3 GB300 disagg sweep — refreshed recipe set (no Marlin variants).
-# All prefill DEP2 (TP1 DP2 EP, 2 GPU/worker). Decode: DEP4, TEP8, DEP8, TEP4.
-# 4 GPU/node (GB300 NVL72). kv-cache-dtype=fp8. srun_options mem=0 required.
-minimaxm3-fp8-gb300-dynamo-vllm-mtp:
-  image: vllm/vllm-openai:nightly-5e35a6f4f9bbc217c599692157ca985c894373f7
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: gb300-nv
-  precision: fp8
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.4.0.dev20260730" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - spec-decoding: "mtp"
-        conc-list: [1]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-1p1d-dep2-tp4-eagle3-c1-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [8]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-1p1d-dep2-tp4-eagle3-c8-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 4
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [1]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-1p1d-dep2-tp8-eagle3-c1-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [4]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-1p1d-dep2-tp8-eagle3-c4-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [8]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-1p1d-dep2-tp8-eagle3-c8-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 1
-          dp-attn: false
-      - spec-decoding: "mtp"
-        conc-list: [64]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-1p1d-dep2-dep8-eagle3-c64-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [512]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-2p1d-dep2-dep8-eagle3-c512-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [256]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-3p1d-dep2-dep8-eagle3-c256-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [1024]
-        prefill:
-          num-worker: 4
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-4p1d-dep2-dep8-eagle3-c1024-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [2048]
-        prefill:
-          num-worker: 4
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-4p1d-dep2-dep8-eagle3-c2048-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-      - spec-decoding: "mtp"
-        conc-list: [2048]
-        prefill:
-          num-worker: 6
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-6p1d-dep2-dep8-eagle3-c2048-mtp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-
-minimaxm3-fp8-gb300-dynamo-vllm:
-  image: vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: gb300-nv
-  precision: fp8
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.3.0.dev20260614" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 1p1d DEP2+DEP8, 3n: conc 256
-      - conc-list: [256]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-1p1d-dep2-dep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-      # 1p1d DEP2+TEP8, 3n: conc 128
-      - conc-list: [128]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-1p1d-dep2-tep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: false
-
-      # 1p2d DEP2+TEP8, 5n: conc 32,64,128
-      - conc-list: [32, 64, 128]
-        prefill:
-          num-worker: 1
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-1p2d-dep2-tep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: false
-
-      # 2p1d DEP2+DEP8, 3n: conc 512
-      - conc-list: [512]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-2p1d-dep2-dep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-      # 2p2d DEP2+TEP8, 5n: conc 16
-      - conc-list: [16]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-2p2d-dep2-tep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: false
-
-      # 2p4d DEP2+TEP4, 5n: conc 4
-      - conc-list: [4]
-        prefill:
-          num-worker: 2
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-2p4d-dep2-tep4-stp.yaml"
-        decode:
-          num-worker: 4
-          tp: 4
-          ep: 4
-          dp-attn: false
-
-      # 3p1d DEP2+DEP8, 4n: conc 1024
-      - conc-list: [1024]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-3p1d-dep2-dep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-      # 3p1d DEP2+DEP16, 6n: conc 512
-      - conc-list: [512]
-        prefill:
-          num-worker: 3
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-3p1d-dep2-dep16-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-      # 6p1d DEP2+DEP8, 5n: conc 2048
-      - conc-list: [2048]
-        prefill:
-          num-worker: 6
-          tp: 2
-          ep: 2
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb300-fp8/8k1k/disagg-6p1d-dep2-dep8-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 8
-          ep: 8
-          dp-attn: true
-
-# MiniMax-M3 GB200 disagg sweep — adapted from NV B300 PR #1863.
-# All prefill DEP4 (TP1 DP4 EP, 4 GPU/worker). Decode: Marlin, TEP8, DEP8, TEP4.
-# 4 GPU/node (GB200 NVL72). FLASHINFER attention with FP8 KV cache.
-minimaxm3-fp8-gb200-dynamo-vllm:
-  image: vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: gb200
-  precision: fp8
-  framework: dynamo-vllm
-  router: { name: dynamo-router, version: "1.3.0.dev20260614" }
-  kv-p2p-transfer: nixl
-  multinode: true
-  disagg: true
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      # 1p2d DEP4+DEP8, 5n: conc 512
-      - conc-list: [512]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb200-fp8/8k1k/disagg-1p1d-dep4-dep16-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-      # 1p2d DEP4+TEP4, 3n: conc 4
-      - conc-list: [4]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb200-fp8/8k1k/disagg-1p2d-dep4-tep4-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: false
-
-      # 1p2d DEP4+TEP8, 5n: conc 4,16,32,64,128
-      - conc-list: [4, 16, 32, 64, 128]
-        prefill:
-          num-worker: 1
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb200-fp8/8k1k/disagg-1p2d-dep4-tep8-stp.yaml"
-        decode:
-          num-worker: 2
-          tp: 8
-          ep: 8
-          dp-attn: false
-
-      # 2p2d DEP4+DEP8, 6n: conc 512,1024
-      - conc-list: [512, 1024]
-        prefill:
-          num-worker: 2
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb200-fp8/8k1k/disagg-2p1d-dep4-dep16-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-      # 3p2d DEP4+DEP8, 7n: conc 4096
-      - conc-list: [4096]
-        prefill:
-          num-worker: 3
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb200-fp8/8k1k/disagg-3p1d-dep4-dep16-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-      # 5p2d DEP4+DEP8, 9n: conc 4096
-      - conc-list: [4096]
-        prefill:
-          num-worker: 5
-          tp: 4
-          ep: 4
-          dp-attn: true
-          additional-settings:
-          - "CONFIG_FILE=recipes/minimaxm3/vllm/gb200-fp8/8k1k/disagg-5p1d-dep4-dep16-stp.yaml"
-        decode:
-          num-worker: 1
-          tp: 16
-          ep: 16
-          dp-attn: true
-
-# MiniMax-M3 day-zero (https://recipes.vllm.ai/MiniMaxAI/MiniMax-M3).
-# 427B total / 26B active MoE with MSA sparse attention; MXFP8 checkpoint
-# (MiniMaxAI/MiniMax-M3-MXFP8, ~444 GB) quantized by NVIDIA — native MX tensor
-# cores on Blackwell. M3 support has not shipped in a stable vLLM release;
-# the dedicated vllm/vllm-openai:minimax-m3-0618-x86_64-cu130 image is built
-# from the m3_release branch (vllm-project/vllm#45381). --block-size 128 is mandatory (MSA
-# sparse/index cache alignment). Weights are NOT SRE-staged; b300 falls back
-# to writable /data/models (see launch_b300-nv.sh MODEL_PATH split).
-minimaxm3-fp8-b300-vllm:
-  image: vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: b300
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 64 }
-      - { tp: 8, ep: 8, conc-start: 1, conc-end: 512 }
-      - { tp: 4, conc-start: 1, conc-end: 128 }
-      - { tp: 4, ep: 4, conc-start: 64, conc-end: 256 }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 64, conc-end: 128 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 128, conc-end: 512 }
-
-# MiniMax-M3 NVFP4 (nvidia/MiniMax-M3-NVFP4) B300 single-node vLLM — FP4 variant
-# of minimaxm3-fp8-b300-vllm. MiniMax-M3 modelopt NVFP4 support (vllm-project/vllm
-# PR #46380) is baked into the perf container image, so no runtime patch is
-# needed. --block-size 128 is mandatory (MSA sparse/index cache);
-# weights are pre-staged read-only at /scratch/models/MiniMax-M3-NVFP4 (added to
-# the STAGED_MODELS allow-list in launch_b300-nv.sh).
-minimaxm3-fp4-b300-vllm:
-  image: vllm/vllm-openai:nightly-5e35a6f4f9bbc217c599692157ca985c894373f7
-  model: nvidia/MiniMax-M3-NVFP4
-  model-prefix: minimaxm3
-  runner: b300
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 2 }
-      - { tp: 4, conc-start: 1, conc-end: 2 }
-      - { tp: 2, conc-start: 4, conc-end: 256 }
-      - { tp: 2, ep: 2, dp-attn: true, conc-start: 512, conc-end: 512 }
-
-# EAGLE3 speculative-decoding (spec-decoding: mtp) variant of MiniMax-M3 NVFP4
-# (nvidia/MiniMax-M3-NVFP4) B300 single-node vLLM, pairing the target with the
-# Inferact/MiniMax-M3-EAGLE3 draft head (3 speculative tokens). MiniMax-M3
-# modelopt NVFP4 support (vllm-project/vllm PR #46380) is baked into the perf
-# container image, so no runtime patch is needed; prompts are routed through the
-# chat template. Target weights are pre-staged read-only at
-# /scratch/models/MiniMax-M3-NVFP4 (added to the STAGED_MODELS allow-list in
-# launch_b300-nv.sh); the EAGLE3 draft is downloaded to the writable models dir.
-minimaxm3-fp4-b300-vllm-mtp:
-  image: vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9
-  model: nvidia/MiniMax-M3-NVFP4
-  model-prefix: minimaxm3
-  runner: b300
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 2, conc-list: [1, 2, 4, 8, 16, 32, 128, 256], spec-decoding: mtp }
-      - { tp: 2, ep: 2, conc-list: [1, 2, 4, 8, 32, 64, 128], spec-decoding: mtp }
-      - { tp: 2, ep: 2, dp-attn: true, conc-list: [128], spec-decoding: mtp }
-      - { tp: 4, conc-list: [1, 2, 4, 8], spec-decoding: mtp }
-      - { tp: 4, ep: 4, conc-list: [1, 2, 4, 8], spec-decoding: mtp }
-      - { tp: 8, conc-list: [1, 2, 4, 8], spec-decoding: mtp }
-
-# MiniMax-M3 day-zero (https://recipes.vllm.ai/MiniMaxAI/MiniMax-M3).
-# 427B total / 26B active MoE with MSA sparse attention; MXFP8 checkpoint
-# (MiniMaxAI/MiniMax-M3-MXFP8, ~444 GB) quantized by NVIDIA — native MX tensor
-# cores on Blackwell. M3 support has not shipped in a stable vLLM release;
-# the dedicated vllm/vllm-openai:minimax-m3-0618-x86_64-cu130 image is built
-# from the m3_release branch (vllm-project/vllm#45381). --block-size 128 is mandatory (MSA
-# sparse/index cache alignment). Weights are NOT SRE-staged: b200-dgxc reads
-# /lustre/fsw/gharunners/models/MiniMax-M3-MXFP8 (pre-downloaded, see
-# launch_b200-dgxc.sh).
-minimaxm3-fp8-b200-vllm:
-  image: vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: b200-dgxc
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 64 }
-      - { tp: 8, ep: 8, conc-start: 1, conc-end: 256 }
-      - { tp: 4, conc-start: 1, conc-end: 64 }
-      - { tp: 4, ep: 4, conc-start: 64, conc-end: 256 }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 64, conc-end: 128 }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 128, conc-end: 256 }
-
-# MiniMax-M3 NVFP4 (nvidia/MiniMax-M3-NVFP4) B200 single-node vLLM — FP4 variant
-# of minimaxm3-fp8-b200-vllm, running on the b200-dgxc cluster. MiniMax-M3
-# modelopt NVFP4 support (vllm-project/vllm PR #46380) is baked into the perf
-# container image, so no runtime patch is needed. --block-size 128 is mandatory
-# (MSA sparse/index cache); weights are pre-staged at /scratch/fsw/models/MiniMax-M3-NVFP4
-# (launch_b200-dgxc.sh resolves MODEL_PATH for minimaxm3-fp4).
-minimaxm3-fp4-b200-vllm:
-  image: vllm/vllm-openai:nightly-5e35a6f4f9bbc217c599692157ca985c894373f7
-  model: nvidia/MiniMax-M3-NVFP4
-  model-prefix: minimaxm3
-  runner: b200-dgxc
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 4 }
-      - { tp: 4, conc-start: 1, conc-end: 4 }
-      - { tp: 2, conc-start: 1, conc-end: 128 }
-      - { tp: 4, conc-start: 256, conc-end: 1024 }
-      - { tp: 4, ep: 4, dp-attn: true, conc-start: 256, conc-end: 1024 }
-
-# EAGLE3 speculative-decoding (spec-decoding: mtp) variant of
-# minimaxm3-fp8-b200-vllm, pairing MiniMaxAI/MiniMax-M3-MXFP8 with the
-# Inferact/MiniMax-M3-EAGLE3 draft head (3 speculative tokens, drafter pinned
-# to TRITON_ATTN). Search space mirrors the non-MTP entry trimmed at the
-# extreme-concurrency end, identical to the minimaxm3-fp8-b300-vllm-mtp
-# precedent: spec decode pays off at low/mid concurrency while acceptance
-# dilutes in big batches, and the draft weights + draft KV shave headroom.
-minimaxm3-fp8-b200-vllm-mtp:
-  image: vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: b200-dgxc
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 8, ep: 8, conc-start: 1, conc-end: 256, spec-decoding: mtp }
-      - { tp: 4, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 4, ep: 4, conc-start: 64, conc-end: 256, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 128, conc-end: 256, spec-decoding: mtp }
-
-# EAGLE3 speculative-decoding (spec-decoding: mtp) variant of MiniMax-M3 NVFP4
-# (nvidia/MiniMax-M3-NVFP4) B200 single-node vLLM, pairing the target with the
-# Inferact/MiniMax-M3-EAGLE3 draft head (3 speculative tokens). Runs on the
-# b200-dgxc cluster. MiniMax-M3 modelopt NVFP4 support (vllm-project/vllm
-# PR #46380) is baked into the perf container image, so no runtime patch is
-# needed; prompts are routed through the chat template. Target weights are
-# pre-staged at /scratch/fsw/models/MiniMax-M3-NVFP4 (launch_b200-dgxc.sh
-# resolves MODEL_PATH for minimaxm3-fp4); the EAGLE3 draft is fetched next to
-# the target weights.
-minimaxm3-fp4-b200-vllm-mtp:
-  image: vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9
-  model: nvidia/MiniMax-M3-NVFP4
-  model-prefix: minimaxm3
-  runner: b200-dgxc
-  precision: fp4
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 2, conc-list: [1, 8, 16, 32, 256], spec-decoding: mtp }
-      - { tp: 2, ep: 2, conc-list: [1, 2, 4, 16, 32, 64, 128, 256, 512], spec-decoding: mtp }
-      - { tp: 4, conc-list: [1, 2, 4, 8, 16, 256], spec-decoding: mtp }
-      - { tp: 4, ep: 4, conc-list: [1, 2, 4, 8, 256], spec-decoding: mtp }
-      - { tp: 8, conc-list: [1, 2, 4, 8], spec-decoding: mtp }
-
-# EAGLE3 speculative-decoding (spec-decoding: mtp) variant of
-# minimaxm3-fp8-b300-vllm, pairing MiniMaxAI/MiniMax-M3-MXFP8 with the
-# Inferact/MiniMax-M3-EAGLE3 draft head (3 speculative tokens).
-# Search space mirrors the non-MTP entry trimmed at
-# the extreme-concurrency end, per the dsv4-fp4-b300-vllm-mtp precedent:
-# spec decode pays off at low/mid concurrency while acceptance dilutes in
-# big batches, and the draft weights + draft KV shave headroom — tp2-ep2 is
-# dropped entirely since its KV headroom was already thin without a draft.
-minimaxm3-fp8-b300-vllm-mtp:
-  image: vllm/vllm-openai:nightly-4080263bb2c5d10deac17aaeb88e0823bc35bca9
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: b300
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 8, ep: 8, conc-start: 1, conc-end: 256, spec-decoding: mtp }
-      - { tp: 4, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 4, ep: 4, conc-start: 64, conc-end: 256, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 128, conc-end: 256, spec-decoding: mtp }
-
-# EAGLE3 speculative-decoding (spec-decoding: mtp) variant of
-# minimaxm3-fp8-h200-vllm (PR #1731), pairing MiniMaxAI/MiniMax-M3-MXFP8 with
-# the Inferact/MiniMax-M3-EAGLE3 draft head (3 speculative tokens, drafter
-# pinned to FLASH_ATTN since the head is MHA and FlashInfer needs GQA/MQA at
-# page size 128). Search space mirrors the non-MTP entry trimmed at the
-# extreme-concurrency end, per the dsv4-fp4-b300-vllm-mtp / minimaxm3 b300-mtp
-# precedent: spec decode pays off at low/mid concurrency while acceptance
-# dilutes in big batches, and the draft weights + draft KV shave headroom.
-minimaxm3-fp8-h200-vllm-mtp:
-  image: vllm/vllm-openai:minimax-m3
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: h200
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-start: 1, conc-end: 32, spec-decoding: mtp }
-      - { tp: 8, conc-start: 1, conc-end: 128, spec-decoding: mtp }
-      - { tp: 8, ep: 8, conc-start: 256, conc-end: 256, spec-decoding: mtp }
-      - { tp: 8, ep: 8, dp-attn: true, conc-start: 256, conc-end: 256, spec-decoding: mtp }
-
-# EAGLE3 speculative-decoding (spec-decoding: mtp) variant of
-# minimaxm3-fp8-h100-vllm (PR #1731). Same TP8-only layout (H100 80 GB has no
-# room below TP8 once ~56 GB of weights land per GPU) plus the
-# Inferact/MiniMax-M3-EAGLE3 draft head (3 speculative tokens, FLASH_ATTN
-# drafter). DEP stays omitted as on the non-MTP entry — KV-cache init already
-# failed at high conc, and the draft head + draft KV only tighten it further.
-minimaxm3-fp8-h100-vllm-mtp:
-  image: vllm/vllm-openai:minimax-m3
-  model: MiniMaxAI/MiniMax-M3-MXFP8
-  model-prefix: minimaxm3
-  runner: h100
-  precision: fp8
-  framework: vllm
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, conc-start: 1, conc-end: 64, spec-decoding: mtp }
-      - { tp: 8, ep: 8, conc-start: 128, conc-end: 256, spec-decoding: mtp }
-
-# Archive source: nvidia-qwen3.5-bf16-master.yaml
-# Deprecated Qwen3.5-397B-A17B bf16 entries archived from nvidia-master.yaml.
-# All bf16 recipes, in every scenario, were deprecated after Monday,
-# August 3, 2026 (see the Deprecation Notice in MODELS.md). fp8 and fp4
-# recipes remain active.
-# Removed from the active master config so sweep generation no longer selects them.
-
-
-qwen3.5-bf16-b200-sglang:
-  image: lmsysorg/sglang:v0.5.14-cu130
-  model: Qwen/Qwen3.5-397B-A17B
-  model-prefix: qwen3.5
-  runner: b200
-  precision: bf16
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64 }
-
-qwen3.5-bf16-b200-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.14-cu130
-  model: Qwen/Qwen3.5-397B-A17B
-  model-prefix: qwen3.5
-  runner: b200
-  precision: bf16
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-
-qwen3.5-bf16-b300-sglang:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: Qwen/Qwen3.5-397B-A17B
-  model-prefix: qwen3.5
-  runner: b300
-  precision: bf16
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64 }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 64 }
-
-qwen3.5-bf16-b300-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.12-cu130
-  model: Qwen/Qwen3.5-397B-A17B
-  model-prefix: qwen3.5
-  runner: b300
-  precision: bf16
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 8, ep: 1, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-      - { tp: 4, ep: 1, conc-start: 4, conc-end: 64, spec-decoding: mtp }
-
-# Retired on 2026-09-22: Docker fixed-sequence coverage is outside the SRT-only cutover.
-# Qwen3.5-397B-A17B NVFP4 single-node SGLang sweep using 4 of 8 RTX PRO
-# 6000 Blackwell GPUs. Both arms use ordinary NCCL collectives on PCIe.
-qwen3.5-fp4-rtx6000pro-sglang:
-  image: lmsysorg/sglang:v0.5.16-cu130
-  model: nvidia/Qwen3.5-397B-A17B-NVFP4
-  model-prefix: qwen3.5
-  runner: rtx6000pro-lat
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-list: [1, 4, 16, 64] }
-      - { tp: 4, ep: 4, conc-list: [1, 4, 16, 64] }
-
-# Same sweep with the built-in MTP draft head driven through SGLang's EAGLE
-# speculative path.
-qwen3.5-fp4-rtx6000pro-sglang-mtp:
-  image: lmsysorg/sglang:v0.5.16-cu130
-  model: nvidia/Qwen3.5-397B-A17B-NVFP4
-  model-prefix: qwen3.5
-  runner: rtx6000pro-lat
-  precision: fp4
-  framework: sglang
-  multinode: false
-  scenarios:
-    fixed-seq-len:
-    - isl: 8192
-      osl: 1024
-      search-space:
-      - { tp: 4, conc-list: [1, 4, 16, 64], spec-decoding: mtp }
-      - { tp: 4, ep: 4, conc-list: [1, 4, 16, 64], spec-decoding: mtp }
diff --git a/docs/configuration-procedures.md b/docs/configuration-procedures.md
index 984c380650..47762f5c94 100644
--- a/docs/configuration-procedures.md
+++ b/docs/configuration-procedures.md
@@ -21,7 +21,7 @@ Use this page for benchmark configuration, recipe, image, and runner changes. It
 | [`perf-changelog.yaml`](../perf-changelog.yaml) | Append-only benchmark trigger log |
 | [`AGENTS.md`](../AGENTS.md) | Repository-wide config, MTP, changelog, and sweep rules |
 
-Archive deprecated entries in [`configs/deprecated/amd-master.yaml`](../configs/deprecated/amd-master.yaml) or [`configs/deprecated/nvidia-master.yaml`](../configs/deprecated/nvidia-master.yaml). Use only these two vendor archives, not separate files per deprecation. Preserve historical settings and comments; disambiguate colliding keys with a descriptive suffix and an original-key comment. For partial retirements, move only the retired scenarios. Keep archives out of active sweep inputs. Retired AMD server-registry entries and model-specific setup belong in `benchmarks/multi_node/amd_utils/deprecated/`, outside the active server lookup. Preserve shared dependencies needed by retained SPEED-Bench collectors, including their scheduling scores. See the [deprecation rules](../AGENTS.md#deprecating-benchmark-configs).
+Delete retired entries from the active master configs; they are not archived. Git history and `perf-changelog.yaml` keep the historical settings. For partial retirements, remove only the retired scenarios. Retired AMD server-registry entries and model-specific setup belong in `benchmarks/multi_node/amd_utils/deprecated/`, outside the active server lookup. Preserve shared dependencies needed by retained SPEED-Bench collectors, including their scheduling scores. See the [deprecation rules](../AGENTS.md#deprecating-benchmark-configs).
 
 ## Dependency submodules
 
@@ -132,11 +132,11 @@ Detailed source: [`.claude/commands/add-model-hardware.md`](../.claude/commands/
 STP (Single Token Prediction) is vanilla autoregressive decoding with one token per forward pass. MTP (Multi-Token Prediction) predicts multiple tokens per forward pass through native heads or speculative decoding.
 
 1. **Fix the identity.** Confirm the exact checkpoint ID, model prefix, precision, architecture, native context, target SKU, framework, and whether decoding is STP, native MTP, or draft-model speculation. Verify the image tag exists. Never invent one.
-2. **Choose two kinds of sibling.** Read the same model on another SKU and another model on the target SKU. Also read the target [`runners/launch_*.sh`](../runners/) and shared [`benchmark_lib.sh`](../benchmarks/benchmark_lib.sh).
-3. **Add the runtime script.** Put the single-node script under [`benchmarks/single_node/fixed_seq_len/`](../benchmarks/single_node/fixed_seq_len/). Preserve the proven sibling's env propagation, parser flags, attention/MoE backend, KV-cache dtype, graph/eager mode, cache setup, and context handling.
+2. **Choose two kinds of sibling.** Read the same model on another SKU and another model on the target SKU. Read their srt-slurm recipes under [`benchmarks/single_node/srt-slurm-recipes/`](../benchmarks/single_node/srt-slurm-recipes/) and master-config entries.
+3. **Add the srt-slurm recipe.** Put it at `benchmarks/single_node/srt-slurm-recipes/<model-prefix>/<engine>/<sku>-<precision>[-mtp]/8k1k.yaml`, with one `override_*` variant per matrix point. Preserve the proven sibling's engine args, env, parser flags, attention/MoE backend, KV-cache dtype, graph/eager mode, `setup_script`, and context handling.
 4. **Add the master entry.** Use [`amd-master.yaml`](../configs/amd-master.yaml) for `mi*`. Otherwise, use [`nvidia-master.yaml`](../configs/nvidia-master.yaml). Set exact `image`, `model`, `model-prefix`, `runner`, `precision`, `framework`, scenarios, and supported search spaces.
 5. **Size from evidence.** Mirror proven parallelism layouts and trim unsupported ones. Latency TP rows normally start at concurrency 1. Do not copy large-memory TP/EP layouts onto a smaller SKU.
-6. **Check launcher routing.** The launcher must resolve the new filename, including framework and `_mtp` suffixes. Simulate STP and MTP resolution and confirm each selected file exists.
+6. **Check variant selection.** Every search-space row carries `srt-recipe:`, and each matrix point must match exactly one recipe variant by TP/GPU count, `CONC`, `KV_OFFLOADING` and image (`infx/srt_slurm/single_node.py::select_recipe`). No launcher routing is needed.
 7. **Append one changelog entry** for the exact new key. See [Append the changelog safely](#append-the-changelog-safely).
 8. **Validate syntax and generated output.** Inspect image, model, runner, ISL/OSL, `max-model-len`, concurrency, TP/PP/EP/DCP/PCP, and `spec-decoding`.
 
@@ -248,7 +248,7 @@ Sources: [`AGENTS.md#non-negotiable-benchmark-invariants`](../AGENTS.md#non-nego
 
 ## Add or change MTP
 
-Sources: [`AGENTS.md#non-negotiable-benchmark-invariants`](../AGENTS.md#non-negotiable-benchmark-invariants), [MTP appendix in the model+hardware playbook](../.claude/commands/add-model-hardware.md#appendix--mtp--eagle3-spec-decoding-variant), and current [`*_mtp.sh` siblings](../benchmarks/single_node/fixed_seq_len/).
+Sources: [`AGENTS.md#non-negotiable-benchmark-invariants`](../AGENTS.md#non-negotiable-benchmark-invariants), [MTP appendix in the model+hardware playbook](../.claude/commands/add-model-hardware.md#appendix--mtp--eagle3-spec-decoding-variant), and current [`*-mtp` srt-slurm recipes](../benchmarks/single_node/srt-slurm-recipes/).
 
 1. Confirm native MTP modules versus an external draft. For a draft, verify exact model ID, method (for example `eagle3`), and recommended speculative-token count from the model/upstream recipe.
 2. Copy a working sibling for the same model and backend. Preserve its speculative config, attention backend, token count, model patches, and dependency setup.
diff --git a/docs/configuration-procedures_zh.md b/docs/configuration-procedures_zh.md
index 83bf44e61c..d4ec6f1a38 100644
--- a/docs/configuration-procedures_zh.md
+++ b/docs/configuration-procedures_zh.md
@@ -21,7 +21,7 @@
 | [`perf-changelog.yaml`](../perf-changelog.yaml) | 只允许追加的基准触发日志 |
 | [`AGENTS.md`](../AGENTS.md) | 仓库级配置、MTP、changelog 和 sweep 规则 |
 
-弃用的配置项应归档至 [`configs/deprecated/amd-master.yaml`](../configs/deprecated/amd-master.yaml) 或 [`configs/deprecated/nvidia-master.yaml`](../configs/deprecated/nvidia-master.yaml)。仅使用这两个按厂商划分的归档文件，不要为每次弃用创建单独文件。保留历史设置和注释；遇到配置键冲突时，使用描述性后缀区分，并用注释记录原始配置键。仅弃用部分场景时，只移出已退役的场景。归档文件不得加入启用的 sweep 输入。退役的 AMD 服务注册项和模型专用初始化逻辑应移至 `benchmarks/multi_node/amd_utils/deprecated/`，不参与启用服务的查找。保留 SPEED-Bench 采集器仍需使用的共享依赖，包括调度评分。参见[弃用规则](../AGENTS.md#deprecating-benchmark-configs)。
+退役的配置项直接从启用的主配置中删除，不再归档；历史设置由 Git 历史和 `perf-changelog.yaml` 保留。仅弃用部分场景时，只删除已退役的场景。退役的 AMD 服务注册项和模型专用初始化逻辑应移至 `benchmarks/multi_node/amd_utils/deprecated/`，不参与启用服务的查找。保留 SPEED-Bench 采集器仍需使用的共享依赖，包括调度评分。参见[弃用规则](../AGENTS.md#deprecating-benchmark-configs)。
 
 ## 依赖子模块
 
@@ -83,11 +83,11 @@ git status --short --branch
 STP（Single Token Prediction，单 Token 预测）是每次前向传播生成一个 Token 的标准自回归解码。MTP（Multi-Token Prediction，多 Token 预测）通过原生预测头或投机解码在每次前向传播中预测多个 Token。
 
 1. **固定身份。**确认精确 checkpoint ID、model prefix、精度、架构、原生上下文、目标 SKU、框架，以及解码方式是 STP、原生 MTP 还是 draft-model 推测。验证镜像 tag 确实存在；绝不能编造。
-2. **选择两类同类项。**阅读同一模型在其他 SKU 上的实现，以及目标 SKU 上的另一个模型。还要阅读目标 [`runners/launch_*.sh`](../runners/) 和共享 [`benchmark_lib.sh`](../benchmarks/benchmark_lib.sh)。
-3. **添加运行时脚本。**把单节点脚本放入 [`benchmarks/single_node/fixed_seq_len/`](../benchmarks/single_node/fixed_seq_len/)。保留已验证同类项中的 env 传递、parser 参数、attention/MoE backend、KV-cache dtype、graph/eager 模式、缓存设置和上下文处理。
+2. **选择两类同类项。**阅读同一模型在其他 SKU 上的实现，以及目标 SKU 上的另一个模型。阅读它们在 [`benchmarks/single_node/srt-slurm-recipes/`](../benchmarks/single_node/srt-slurm-recipes/) 下的 srt-slurm 配方及主配置条目。
+3. **添加 srt-slurm 配方。**放在 `benchmarks/single_node/srt-slurm-recipes/<model-prefix>/<engine>/<sku>-<precision>[-mtp]/8k1k.yaml`，每个矩阵点对应一个 `override_*` 变体。保留已验证同类项中的引擎参数、env、parser 参数、attention/MoE backend、KV-cache dtype、graph/eager 模式、`setup_script` 和上下文处理。
 4. **添加主配置条目。**`mi*` 使用 [`amd-master.yaml`](../configs/amd-master.yaml)，其他使用 [`nvidia-master.yaml`](../configs/nvidia-master.yaml)。精确设置 `image`、`model`、`model-prefix`、`runner`、`precision`、`framework`、scenario 和支持的搜索空间。
 5. **依据证据确定规模。**复制已验证的并行布局并删除不支持的布局。延迟型 TP 行通常从并发 1 开始；不要把大显存 SKU 的 TP/EP 布局复制到小显存 SKU。
-6. **检查 launcher 路由。**launcher 必须能解析新文件名，包括 framework 和 `_mtp` 后缀。模拟 STP 与 MTP 解析，并确认每个被选择的文件都存在。
+6. **检查变体选择。**每个搜索空间行都带有 `srt-recipe:`，每个矩阵点必须按 TP/GPU 数、`CONC`、`KV_OFFLOADING` 和镜像恰好匹配一个配方变体（`infx/srt_slurm/single_node.py::select_recipe`），无需 launcher 路由。
 7. **追加一条 changelog**，精确选择新 key；参见[安全追加 changelog](#安全追加-changelog)。
 8. **验证语法和生成结果。**检查 image、model、runner、ISL/OSL、`max-model-len`、并发、TP/PP/EP/DCP/PCP 和 `spec-decoding`。
 
@@ -197,7 +197,7 @@ llm-d 不是 srt-slurm 路径：InferenceX 自己持有 Slurm allocation，并
 
 ## 添加或修改 MTP
 
-来源：[`AGENTS.md#non-negotiable-benchmark-invariants`](../AGENTS.md#non-negotiable-benchmark-invariants)、[模型+硬件 playbook 的 MTP 附录](../.claude/commands/add-model-hardware.md#appendix--mtp--eagle3-spec-decoding-variant)和现有 [`*_mtp.sh` 同类项](../benchmarks/single_node/fixed_seq_len/)。
+来源：[`AGENTS.md#non-negotiable-benchmark-invariants`](../AGENTS.md#non-negotiable-benchmark-invariants)、[模型+硬件 playbook 的 MTP 附录](../.claude/commands/add-model-hardware.md#appendix--mtp--eagle3-spec-decoding-variant)和现有 [`*-mtp` srt-slurm 配方](../benchmarks/single_node/srt-slurm-recipes/)。
 
 1. 确认使用原生 MTP 模块还是外部 draft。使用 draft 时，从模型/上游配方验证精确模型 ID、方法（例如 `eagle3`）和建议 speculative token 数。
 2. 复制相同模型和 backend 的可工作同类项。保留其 speculative config、attention backend、token 数、模型补丁和依赖设置。

From 4f319cb15955e370863288522c8814b690dc39a9 Mon Sep 17 00:00:00 2001
From: functionstackx <47992694+functionstackx@users.noreply.github.com>
Date: Sat, 26 Sep 2026 02:41:24 -0400
Subject: [PATCH 5/6] Delete benchmarks/multi_node/amd_utils/deprecated

Nothing loads these archived AMD server-registry files or glm5_setup.sh;
the live registries are the models*.yaml and node_excludes.yaml beside
them. configuration-procedures now says retired AMD registry entries are
deleted rather than moved into deprecated/.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
---
 .../amd_utils/deprecated/glm5_setup.sh        | 30 --------
 .../amd_utils/deprecated/models.yaml          | 69 -------------------
 .../amd_utils/deprecated/models_atom.yaml     | 57 ---------------
 .../amd_utils/deprecated/models_vllm.yaml     | 47 -------------
 .../amd_utils/deprecated/node_excludes.yaml   | 14 ----
 docs/configuration-procedures.md              |  2 +-
 docs/configuration-procedures_zh.md           |  2 +-
 7 files changed, 2 insertions(+), 219 deletions(-)
 delete mode 100644 benchmarks/multi_node/amd_utils/deprecated/glm5_setup.sh
 delete mode 100644 benchmarks/multi_node/amd_utils/deprecated/models.yaml
 delete mode 100644 benchmarks/multi_node/amd_utils/deprecated/models_atom.yaml
 delete mode 100644 benchmarks/multi_node/amd_utils/deprecated/models_vllm.yaml
 delete mode 100644 benchmarks/multi_node/amd_utils/deprecated/node_excludes.yaml

diff --git a/benchmarks/multi_node/amd_utils/deprecated/glm5_setup.sh b/benchmarks/multi_node/amd_utils/deprecated/glm5_setup.sh
deleted file mode 100644
index 94b05edff0..0000000000
--- a/benchmarks/multi_node/amd_utils/deprecated/glm5_setup.sh
+++ /dev/null
@@ -1,30 +0,0 @@
-#!/bin/bash
-
-# Historical GLM-5 setup, retained for explicit replay only.
-# Source after the shared setup helper when restoring the archived GLM-5 registry.
-# GLM-5 needs a transformers build with the glm_moe_dsa model type, which the mori
-# images do not ship. Gated on any GLM model name.
-install_transformers_glm5() {
-    if [[ "$MODEL_NAME" != *GLM* ]]; then
-        return 0
-    fi
-
-    if python3 -c "from transformers import AutoConfig; AutoConfig.from_pretrained('zai-org/GLM-5-FP8', trust_remote_code=True)" 2>/dev/null; then
-        echo "[SETUP] transformers already supports GLM-5 model type"
-        return 0
-    fi
-
-    echo "[SETUP] Installing transformers with GLM-5 (glm_moe_dsa) support..."
-    pip install --quiet -U --no-cache-dir \
-        "git+https://github.com/huggingface/transformers.git@6ed9ee36f608fd145168377345bfc4a5de12e1e2"
-    _SETUP_INSTALLED+=("transformers-glm5")
-}
-
-install_transformers_glm5
-
-# Historical GLM-5 environment overrides.
-if [[ "$MODEL_NAME" == "GLM-5-FP8" ]]; then
-    export SGLANG_ROCM_FUSED_DECODE_MLA=0
-    export ROCM_QUICK_REDUCE_QUANTIZATION=INT4
-    export SAFETENSORS_FAST_GPU=1
-fi
diff --git a/benchmarks/multi_node/amd_utils/deprecated/models.yaml b/benchmarks/multi_node/amd_utils/deprecated/models.yaml
deleted file mode 100644
index d6dfd0d39c..0000000000
--- a/benchmarks/multi_node/amd_utils/deprecated/models.yaml
+++ /dev/null
@@ -1,69 +0,0 @@
-# Historical server registry entries; excluded from active server lookup.
-
-GLM-5-FP8:
-  base_flags: "--decode-log-interval 1000 --log-level warning --watchdog-timeout 3600 --load-balance-method round_robin --disaggregation-transfer-backend mori --tool-call-parser glm47 --reasoning-parser glm45 --model-loader-extra-config '{\\\"enable_multithread_load\\\": true, \\\"num_threads\\\": 8}'"
-  mtp_flags: ""
-  dp_flags: "--enable-dp-attention --moe-dense-tp-size 1 --enable-dp-lm-head"
-  ep_flags: "--moe-a2a-backend mori"
-  prefill:
-    mem_fraction_static: 0.8
-    disable_radix_cache: true
-    dp:
-      max_running_requests: 24
-      chunked_prefill_size: "MORI_MAX_DISPATCH_TOKENS_PREFILL * PREFILL_TP_SIZE"
-      cuda_graph_bs: "1 2 3"
-    no_dp:
-      max_running_requests: 128
-      chunked_prefill_size: 262144
-      cuda_graph_bs_range: "1-128"
-  decode:
-    mem_fraction_static: 0.85
-    dp:
-      max_running_requests: 4096
-      chunked_prefill_size: "MORI_MAX_DISPATCH_TOKENS_DECODE * DECODE_TP_SIZE"
-      cuda_graph_bs_range: "1-160"
-    ep_only:
-      max_running_requests: 256
-      chunked_prefill_size: 262144
-      cuda_graph_bs_range: "1-256"
-    no_dp:
-      max_running_requests: 128
-      chunked_prefill_size: 262144
-      cuda_graph_bs_range: "1-128"
-
-
-DeepSeek-V4-Pro-DI:
-  base_flags: "--decode-log-interval 100 --log-level info --watchdog-timeout 3600 --load-balance-method round_robin --kv-cache-dtype fp8_e4m3 --attention-backend dsv4 --page-size 256 --swa-full-tokens-ratio 0.1 --disable-shared-experts-fusion --tool-call-parser deepseekv4 --reasoning-parser deepseek-v4 --disaggregation-transfer-backend mori"
-  # DeepSeek-V4 built-in MTP: load via the EAGLE spec path with eagle-topk 1 (a single
-  # MTP chain). NOT NEXTN -- the V3/R1 NEXTN loader crashes the dsv4 decode server at
-  # init on the V4 architecture. Mirrors the sglang dsv4 recipes
-  # (srt-slurm-recipes/dsv4/sglang/*: EAGLE + eagle-topk 1, in-checkpoint draft,
-  # no draft-model-path). Consumed by build_server_config only when DECODE_MTP_SIZE > 0
-  # (spec-decoding: mtp); a no-op for the spec-none base scenario (DECODE_MTP_SIZE=0).
-  # build_server_config appends --speculative-num-steps/--speculative-num-draft-tokens
-  # from DECODE_MTP_SIZE.
-  mtp_flags: "--speculative-algorithm EAGLE --speculative-eagle-topk 1"
-  dp_flags: "--enable-dp-attention --moe-dense-tp-size 1 --enable-dp-lm-head"
-  ep_flags: "--ep-dispatch-algorithm fake --moe-a2a-backend mori --deepep-mode normal"
-  prefill:
-    mem_fraction_static: 0.8
-    disable_radix_cache: true
-    disable_cuda_graph: true
-    dp:
-      max_running_requests: 1024
-      chunked_prefill_size: 131072
-      context_length: 9217
-      max_total_tokens: 2097152
-    no_dp:
-      max_running_requests: 128
-      chunked_prefill_size: 131072
-      context_length: 9217
-      max_total_tokens: 2097152
-  decode:
-    mem_fraction_static: 0.85
-    dp:
-      max_running_requests: 1024
-      cuda_graph_bs_range: "1-128"
-    no_dp:
-      max_running_requests: 128
-      cuda_graph_bs_range: "1-128"
diff --git a/benchmarks/multi_node/amd_utils/deprecated/models_atom.yaml b/benchmarks/multi_node/amd_utils/deprecated/models_atom.yaml
deleted file mode 100644
index adaaa090eb..0000000000
--- a/benchmarks/multi_node/amd_utils/deprecated/models_atom.yaml
+++ /dev/null
@@ -1,57 +0,0 @@
-# Historical server registry entries; excluded from active server lookup.
-
-Kimi-K2.5-MXFP4:
-  env: "AITER_QUICK_REDUCE_QUANTIZATION=INT4 AITER_MXFP4_INTERMEDIATE=1"
-  kv_cache_flags: "--kv_cache_dtype fp8"
-  mem_frac_static: "0.9"
-  max_model_len: "32768"
-  max_num_seqs: "256"
-  max_num_batched_tokens: "32768"
-
-DeepSeek-V4-Pro:
-  env: "ATOM_MOE_GU_ITLV=1 AITER_BF16_FP8_MOE_BOUND=0"
-  kv_cache_flags: "--kv_cache_dtype fp8"
-  tp_dp_flags: "--enable-dp-attention --enable-tbo"
-  prefill_tp_dp_flags: "--enable-dp-attention --enable-tbo"
-  decode_tp_dp_flags: "--enable-dp-attention --enable-tbo"
-  tp_dp_env: "GPU_MAX_HW_QUEUES=5 ATOM_CPU_AFFINITY=1"
-  ep_dp_flags: "--enable-expert-parallel --enable-dp-attention"
-  prefill_ep_dp_flags: "--enable-expert-parallel --enable-dp-attention"
-  decode_ep_dp_flags: "--enable-expert-parallel --enable-dp-attention"
-  mtp_flags: "--method mtp --num-speculative-tokens"
-
-MiniMax-M3-MXFP4:
-  env: "AITER_QUICK_REDUCE_QUANTIZATION=INT4 ATOM_FORCE_ATTN_TRITON=1"
-  kv_cache_flags: "--kv_cache_dtype fp8"
-  tp_dp_flags: "--enable-dp-attention"
-  prefill_tp_dp_flags: "--enable-dp-attention --enable-tbo prefill"
-  decode_tp_dp_flags: "--enable-dp-attention"
-  ep_dp_flags: "--enable-expert-parallel --enable-dp-attention"
-  prefill_ep_dp_flags: "--enable-expert-parallel --enable-dp-attention"
-  decode_ep_dp_flags: "--enable-expert-parallel --enable-dp-attention"
-  mtp_flags: "--method eagle3 --draft-model Inferact/MiniMax-M3-EAGLE3 --num-speculative-tokens"
-  online_quant_config: '{"global_quant_config":"ptpc_fp8","exclude_layer":["lm_head","model.embed_tokens","vision_tower","multi_modal_projector","patch_merge_mlp","*block_sparse_moe"]}'
-  online_quant_dpa_config: '{"global_quant_config":"ptpc_fp8","exclude_layer":["lm_head","model.embed_tokens","vision_tower","multi_modal_projector","patch_merge_mlp","*block_sparse_moe"]}'
-  block_size: "128"
-  mem_frac_static: "0.8"
-  max_model_len: "32768"
-  max_num_seqs: "256"
-  max_num_batched_tokens: "32768"
-
-MiniMax-M3-MXFP8:
-  env: "AITER_QUICK_REDUCE_QUANTIZATION=INT4 ATOM_FORCE_ATTN_TRITON=1"
-  kv_cache_flags: "--kv_cache_dtype fp8"
-  tp_dp_flags: "--enable-dp-attention"
-  prefill_tp_dp_flags: "--enable-dp-attention --enable-tbo prefill"
-  decode_tp_dp_flags: "--enable-dp-attention"
-  ep_dp_flags: "--enable-expert-parallel --enable-dp-attention"
-  prefill_ep_dp_flags: "--enable-expert-parallel --enable-dp-attention"
-  decode_ep_dp_flags: "--enable-expert-parallel --enable-dp-attention"
-  mtp_flags: "--method eagle3 --draft-model Inferact/MiniMax-M3-EAGLE3 --num-speculative-tokens"
-  online_quant_config: '{"global_quant_config":"ptpc_fp8","exclude_layer":["lm_head","model.embed_tokens","vision_tower","multi_modal_projector","patch_merge_mlp","*block_sparse_moe"]}'
-  online_quant_dpa_config: '{"global_quant_config":"ptpc_fp8","exclude_layer":["lm_head","model.embed_tokens","vision_tower","multi_modal_projector","patch_merge_mlp","*.gate.*","*.block_sparse_moe.experts*"]}'
-  block_size: "128"
-  mem_frac_static: "0.8"
-  max_model_len: "32768"
-  max_num_seqs: "256"
-  max_num_batched_tokens: "32768"
diff --git a/benchmarks/multi_node/amd_utils/deprecated/models_vllm.yaml b/benchmarks/multi_node/amd_utils/deprecated/models_vllm.yaml
deleted file mode 100644
index f9a54d918a..0000000000
--- a/benchmarks/multi_node/amd_utils/deprecated/models_vllm.yaml
+++ /dev/null
@@ -1,47 +0,0 @@
-# Historical server registry entries; excluded from active server lookup.
-
-Kimi-K2.5-MXFP4:
-  prefill_flags: "--tensor-parallel-size 8 --no-enable-prefix-caching --block-size 1 --gpu-memory-utilization 0.90 --max-model-len 32768 --mm-encoder-tp-mode data --kv-cache-dtype fp8 --max-num-seqs 256 --max-num-batched-tokens 32768"
-  decode_flags: "--tensor-parallel-size 8 --all2all-backend mori_low_latency --no-enable-prefix-caching --block-size 1 --gpu-memory-utilization 0.90 --max-model-len 32768 --mm-encoder-tp-mode data --kv-cache-dtype fp8 --max-num-seqs 256 --max-num-batched-tokens 32768"
-  env: "VLLM_USE_V1=1 VLLM_ROCM_USE_AITER=1 VLLM_ROCM_USE_AITER_RMSNORM=1 VLLM_ROCM_QUICK_REDUCE_QUANTIZATION=INT4 HSA_NO_SCRATCH_RECLAIM=1 VLLM_ENGINE_READY_TIMEOUT_S=3600"
-  hf_dir: "models--amd--Kimi-K2.5-MXFP4"
-
-
-MiniMax-M2.5:
-  # AITER fused-MoE kernel fmoe_bf16_blockscaleFp8_g1u1_vs_silu_32x384 for gfx950 writes OOB when run with MiniMax's shapes at M=8K(=num batched tokens), crashing vllm during AITER warmup.
-  # Set token budget to 4k to avoid using that shape, instead of disabling AITER_MOE.
-  prefill_flags: "--max-num-batched-tokens 4K --tensor-parallel-size 8 --enable-expert-parallel --all2all-backend mori_low_latency --no-enable-prefix-caching --gpu-memory-utilization 0.95 --block-size 32"
-  decode_flags: "--max-num-batched-tokens 4K --tensor-parallel-size 8 --enable-expert-parallel --all2all-backend mori_low_latency --no-enable-prefix-caching --gpu-memory-utilization 0.95 --block-size 32"
-  env: "VLLM_USE_V1=1 VLLM_ROCM_USE_AITER=1 VLLM_ROCM_QUICK_REDUCE_QUANTIZATION=INT4 VLLM_ENGINE_READY_TIMEOUT_S=3600 VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT=1"
-  hf_dir: "models--MiniMaxAI--MiniMax-M2.5"
-
-
-gpt-oss-120b:
-  prefill_flags: "--tensor-parallel-size 8"
-  decode_flags: "--tensor-parallel-size 8"
-  env: "VLLM_USE_V1=1 VLLM_ROCM_USE_AITER=1 VLLM_ROCM_USE_AITER_TRITON_BF16_GEMM=0 VLLM_USE_AITER_UNIFIED_ATTENTION=1 VLLM_ROCM_USE_AITER_MHA=0 ROCM_TRITON_MOE_PRESHUFFLE_SCALES=0"
-
-MiniMax-M3-MXFP4:
-  prefill_flags: "--tensor-parallel-size 8 --max-num-batched-tokens 32768 --max-num-seqs 512 --block-size 128 --language-model-only --attention-backend TRITON_ATTN --moe-backend aiter --no-enable-prefix-caching --gpu-memory-utilization 0.90 --tool-call-parser minimax_m3 --reasoning-parser minimax_m3 --enable-auto-tool-choice"
-  decode_flags: "--tensor-parallel-size 8 --max-num-batched-tokens 32768 --max-num-seqs 512 --block-size 128 --language-model-only --attention-backend TRITON_ATTN --moe-backend aiter --no-enable-prefix-caching --gpu-memory-utilization 0.90 --tool-call-parser minimax_m3 --reasoning-parser minimax_m3 --enable-auto-tool-choice"
-  env: "VLLM_USE_V1=1 VLLM_ROCM_USE_AITER=1 VLLM_ROCM_USE_AITER_MOE=1 VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS=1 VLLM_USE_BREAKABLE_CUDAGRAPH=0 VLLM_ENGINE_READY_TIMEOUT_S=3600"
-  prefill_env: "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION=INT4 VLLM_ROCM_QUICK_REDUCE_MAX_SIZE_BYTES_MB=2048"
-  hf_dir: "models--amd--MiniMax-M3-MXFP4"
-
-MiniMax-M3-MXFP8:
-  # MiniMax-M3 MXFP8 disagg, no EP. The --tensor-parallel-size 8 below is just a
-  # placeholder: server_vllm.sh sed-rewrites it to PREFILL_TP_SIZE/DECODE_TP_SIZE
-  # from the master-config prefill/decode tp (the sweep uses TP4 workers only,
-  # varying the prefill:decode worker ratio rather than TP).
-  # --block-size 128 is mandatory (MSA sparse/index cache); text-only benchmark
-  # so --language-model-only frees the vision encoder. gfx950 uses FP8 KV cache.
-  # Serve flags + env were kept in sync with the former single-node
-  # minimaxm3_fp8_mi355x.sh recipe (since deleted):
-  #   --moe-backend aiter (AITER fused MoE), --linear-backend emulation, and the
-  #   larger --max-num-batched-tokens 32768 prefill budget; plus the AITER
-  #   router-append shared-experts fusion (self-disables under EP; the sweep is
-  #   EP1 so it stays active) and INT6 quick all-reduce quantization.
-  prefill_flags: "--tensor-parallel-size 8 --block-size 128 --language-model-only --kv-cache-dtype fp8 --attention-backend TRITON_ATTN --moe-backend aiter --linear-backend emulation --max-num-batched-tokens 32768 --no-enable-prefix-caching --gpu-memory-utilization 0.90 --tool-call-parser minimax_m3 --reasoning-parser minimax_m3 --enable-auto-tool-choice"
-  decode_flags: "--tensor-parallel-size 8 --block-size 128 --language-model-only --kv-cache-dtype fp8 --attention-backend TRITON_ATTN --moe-backend aiter --linear-backend emulation --max-num-batched-tokens 32768 --no-enable-prefix-caching --gpu-memory-utilization 0.90 --tool-call-parser minimax_m3 --reasoning-parser minimax_m3 --enable-auto-tool-choice"
-  env: "VLLM_USE_V1=1 VLLM_ROCM_USE_AITER=1 VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS=1 VLLM_ROCM_QUICK_REDUCE_QUANTIZATION=INT6 VLLM_USE_BREAKABLE_CUDAGRAPH=0 VLLM_ENGINE_READY_TIMEOUT_S=3600"
-  hf_dir: "models--MiniMaxAI--MiniMax-M3-MXFP8"
diff --git a/benchmarks/multi_node/amd_utils/deprecated/node_excludes.yaml b/benchmarks/multi_node/amd_utils/deprecated/node_excludes.yaml
deleted file mode 100644
index 6ba840328e..0000000000
--- a/benchmarks/multi_node/amd_utils/deprecated/node_excludes.yaml
+++ /dev/null
@@ -1,14 +0,0 @@
-# Per (framework, model) SLURM node exclusions, consumed by submit.sh.
-#
-# A rule applies when FRAMEWORK matches "framework" exactly and MODEL_NAME is
-# in "models"; its exclude_nodes are then passed to sbatch --exclude. Any
-# (framework, model) combination with no matching rule gets an empty exclude
-# list (i.e. no nodes excluded).
-#
-# SLURM_EXCLUDE_NODES env var, if set, always overrides this file.
-
-rules:
-  - framework: vllm-disagg
-    models:
-      - Kimi-K2.5-MXFP4
-    exclude_nodes: mia1-p01-g09,mia1-p01-g10,mia1-p01-g11,mia1-p01-g12
diff --git a/docs/configuration-procedures.md b/docs/configuration-procedures.md
index 47762f5c94..779db276ba 100644
--- a/docs/configuration-procedures.md
+++ b/docs/configuration-procedures.md
@@ -21,7 +21,7 @@ Use this page for benchmark configuration, recipe, image, and runner changes. It
 | [`perf-changelog.yaml`](../perf-changelog.yaml) | Append-only benchmark trigger log |
 | [`AGENTS.md`](../AGENTS.md) | Repository-wide config, MTP, changelog, and sweep rules |
 
-Delete retired entries from the active master configs; they are not archived. Git history and `perf-changelog.yaml` keep the historical settings. For partial retirements, remove only the retired scenarios. Retired AMD server-registry entries and model-specific setup belong in `benchmarks/multi_node/amd_utils/deprecated/`, outside the active server lookup. Preserve shared dependencies needed by retained SPEED-Bench collectors, including their scheduling scores. See the [deprecation rules](../AGENTS.md#deprecating-benchmark-configs).
+Delete retired entries from the active master configs; they are not archived. Git history and `perf-changelog.yaml` keep the historical settings. For partial retirements, remove only the retired scenarios. Delete retired AMD server-registry entries and model-specific setup from `benchmarks/multi_node/amd_utils/` as well. Preserve shared dependencies needed by retained SPEED-Bench collectors, including their scheduling scores. See the [deprecation rules](../AGENTS.md#deprecating-benchmark-configs).
 
 ## Dependency submodules
 
diff --git a/docs/configuration-procedures_zh.md b/docs/configuration-procedures_zh.md
index d4ec6f1a38..6243070908 100644
--- a/docs/configuration-procedures_zh.md
+++ b/docs/configuration-procedures_zh.md
@@ -21,7 +21,7 @@
 | [`perf-changelog.yaml`](../perf-changelog.yaml) | 只允许追加的基准触发日志 |
 | [`AGENTS.md`](../AGENTS.md) | 仓库级配置、MTP、changelog 和 sweep 规则 |
 
-退役的配置项直接从启用的主配置中删除，不再归档；历史设置由 Git 历史和 `perf-changelog.yaml` 保留。仅弃用部分场景时，只删除已退役的场景。退役的 AMD 服务注册项和模型专用初始化逻辑应移至 `benchmarks/multi_node/amd_utils/deprecated/`，不参与启用服务的查找。保留 SPEED-Bench 采集器仍需使用的共享依赖，包括调度评分。参见[弃用规则](../AGENTS.md#deprecating-benchmark-configs)。
+退役的配置项直接从启用的主配置中删除，不再归档；历史设置由 Git 历史和 `perf-changelog.yaml` 保留。仅弃用部分场景时，只删除已退役的场景。退役的 AMD 服务注册项和模型专用初始化逻辑也应从 `benchmarks/multi_node/amd_utils/` 中删除。保留 SPEED-Bench 采集器仍需使用的共享依赖，包括调度评分。参见[弃用规则](../AGENTS.md#deprecating-benchmark-configs)。
 
 ## 依赖子模块
 

From 90c742aaef35cf176dba131a69515c866c387259 Mon Sep 17 00:00:00 2001
From: functionstackx <47992694+functionstackx@users.noreply.github.com>
Date: Sat, 26 Sep 2026 02:46:04 -0400
Subject: [PATCH 6/6] Revert the docs/waiver/2508.md edit

Leave the historical waiver record unchanged.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
---
 docs/waiver/2508.md | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/docs/waiver/2508.md b/docs/waiver/2508.md
index 66e9e9afa4..6fdc3e7c64 100644
--- a/docs/waiver/2508.md
+++ b/docs/waiver/2508.md
@@ -9,9 +9,8 @@ prohibits unless covered by a filled-out waiver.
 - **Master config entry:** `kimik3-fp4-mi355x-vllm-agentic-mtp` in [`configs/amd-master.yaml`](../../configs/amd-master.yaml)
 - **Pinned image:** `vllm/vllm-openai-rocm:nightly-cb8104839c141609d99f1254459ef3a4f1bd4263`
 - **Patch entrypoint:** `bash "$(dirname "$0")/apply_k3_container_patches.sh"` invoked from
-  the legacy `kimik3_fp4_mi355x_mtp.sh` AgentX script
-- **Patch script:** `apply_k3_container_patches.sh`
-- **Status:** both scripts, and the `kimik3-fp4-mi355x-vllm-agentic-mtp` config that used them, were removed with the legacy single-node AgentX bash path (#3460, #3461).
+  [`benchmarks/single_node/agentic/kimik3_fp4_mi355x_mtp.sh`](../../benchmarks/single_node/agentic/kimik3_fp4_mi355x_mtp.sh)
+- **Patch script:** [`benchmarks/single_node/agentic/apply_k3_container_patches.sh`](../../benchmarks/single_node/agentic/apply_k3_container_patches.sh)
 
 ## What is patched