diff --git a/models/alibaba/qwen-audio-3.0-tts-plus.toml b/models/alibaba/qwen-audio-3.0-tts-plus.toml
new file mode 100644
index 00000000000..aae03164ee3
--- /dev/null
+++ b/models/alibaba/qwen-audio-3.0-tts-plus.toml
@@ -0,0 +1,17 @@
+name = "Qwen-Audio-3.0-TTS-Plus"
+description = "Qwen-Audio-3.0-TTS-Plus is Alibaba's higher-quality text-to-speech model, generating spoken audio from text via the DashScope Speech Synthesizer API."
+release_date = "2026-08-05"
+last_updated = "2026-08-05"
+attachment = false
+reasoning = false
+temperature = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 0
+output = 0
+
+[modalities]
+input = ["text"]
+output = ["audio"]
diff --git a/models/alibaba/qwen-image-2.0-pro.toml b/models/alibaba/qwen-image-2.0-pro.toml
new file mode 100644
index 00000000000..285a450761c
--- /dev/null
+++ b/models/alibaba/qwen-image-2.0-pro.toml
@@ -0,0 +1,18 @@
+# https://www.alibabacloud.com/help/en/model-studio/qwen-image-api
+name = "Qwen-Image-2.0-Pro"
+description = "Qwen-Image-2.0-Pro is an image generation model launched by Qwen AI."
+release_date = "2026-03-03"
+last_updated = "2026-03-03"
+attachment = true
+reasoning = false
+temperature = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 1_300
+output = 0
+
+[modalities]
+input = ["text", "image"]
+output = ["image"]
diff --git a/models/alibaba/qwen-image-2.0.toml b/models/alibaba/qwen-image-2.0.toml
new file mode 100644
index 00000000000..9d89f4d8957
--- /dev/null
+++ b/models/alibaba/qwen-image-2.0.toml
@@ -0,0 +1,18 @@
+# https://www.alibabacloud.com/help/en/model-studio/qwen-image-api
+name = "Qwen-Image-2.0"
+description = "Qwen-Image-2.0 is an image generation model launched by Qwen AI."
+release_date = "2026-03-03"
+last_updated = "2026-03-03"
+attachment = true
+reasoning = false
+temperature = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 1_300
+output = 0
+
+[modalities]
+input = ["text", "image"]
+output = ["image"]
diff --git a/models/alibaba/qwen-image-3.0-pro.toml b/models/alibaba/qwen-image-3.0-pro.toml
new file mode 100644
index 00000000000..156aaa994b3
--- /dev/null
+++ b/models/alibaba/qwen-image-3.0-pro.toml
@@ -0,0 +1,18 @@
+# https://www.alibabacloud.com/help/en/model-studio/qwen-image-generation-and-editing-api-reference
+name = "Qwen-Image-3.0-Pro"
+description = "Rich content: Supports input of up to 4.5k tokens and dense information layout with images-within-images, enabling complex layouts like newspapers, storyboards, menus, and exam papers to be generated in a single pass. Authentic detail: Supports precise rendering of text as small as 10px, and vividly reproduces fine details such as micro-expressions, pores, and individual strands of hair—approaching the quality of real photography. Deep knowledge: Supports native rendering of 12 languages and 20+ fonts, realistic simulation of mainstream interfaces such as web pages, games, and live streams, fully incorporating external knowledge. Qwen-Image-3.0-Pro isn't just pursuing \"good looks\"—it's pursuing \"usefulness\", making image generation a truly deployable productivity tool."
+release_date = "2026-08-05"
+last_updated = "2026-08-05"
+attachment = true
+reasoning = false
+temperature = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 4_500
+output = 0
+
+[modalities]
+input = ["text", "image"]
+output = ["image"]
diff --git a/models/alibaba/qwen-image-3.0.toml b/models/alibaba/qwen-image-3.0.toml
new file mode 100644
index 00000000000..84ba68f4c6b
--- /dev/null
+++ b/models/alibaba/qwen-image-3.0.toml
@@ -0,0 +1,18 @@
+# https://www.alibabacloud.com/help/en/model-studio/qwen-image-generation-and-editing-api-reference
+name = "Qwen-Image-3.0"
+description = "Accurate Prompt Understanding: Supports up to 4.5k token inputs, accurately interpreting complex text-and-image prompts and generating dense layouts in one pass. Reliable Text Rendering: Delivers crisp 10px text across 12 languages and 20+ fonts, making infographics and interfaces ready to use. Efficient Batch Production: Scales everyday tasks like posters, web pages, and UI screens with better cost efficiency, keeping creative output sustainable. Qwen-Image-3.0 Standard is built not just for visual quality, but for smooth, reliable daily creation—turning image generation into sustainable content productivity."
+release_date = "2026-08-05"
+last_updated = "2026-08-05"
+attachment = true
+reasoning = false
+temperature = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 4_500
+output = 0
+
+[modalities]
+input = ["text", "image"]
+output = ["image"]
diff --git a/models/alibaba/qwen3-14b.toml b/models/alibaba/qwen3-14b.toml
new file mode 100644
index 00000000000..fd5ef6377ea
--- /dev/null
+++ b/models/alibaba/qwen3-14b.toml
@@ -0,0 +1,24 @@
+name = "Qwen3 14B"
+description = "Qwen instruction model for multilingual chat, reasoning, and tool use"
+family = "qwen"
+release_date = "2025-04-28"
+last_updated = "2025-04-28"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+knowledge = "2025-04"
+open_weights = true
+license = "Apache 2.0"
+
+[limit]
+context = 131_072
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/Qwen/Qwen3-14B"
diff --git a/models/alibaba/qwen3-235b-a22b-thinking-2507.toml b/models/alibaba/qwen3-235b-a22b-thinking-2507.toml
new file mode 100644
index 00000000000..d6ccc62a6e4
--- /dev/null
+++ b/models/alibaba/qwen3-235b-a22b-thinking-2507.toml
@@ -0,0 +1,23 @@
+# https://huggingface.co/Qwen/Qwen3-235B-A22B-Thinking-2507
+name = "Qwen3 235B A22B Thinking 2507"
+description = "Qwen3-235B-A22B-Thinking-2507 is a high-performance, open-weight Mixture-of-Experts (MoE) language model optimized for complex reasoning tasks. It activates 22B of its 235B parameters per forward pass and natively supports up to 262,144 tokens of context. This \"thinking-only\" variant enhances structured logical reasoning, mathematics, science, and long-form generation, showing strong benchmark performance across AIME, SuperGPQA, LiveCodeBench, and MMLU-Redux. It enforces a special reasoning mode (</think>) and is designed for high-token outputs (up to 81,920 tokens) in challenging domains.\n\nThe model is instruction-tuned and excels at step-by-step reasoning, tool use, agentic workflows, and multilingual tasks. This release represents the most capable open-source variant in the Qwen3-235B series, surpassing many closed models in structured reasoning use cases."
+release_date = "2025-07-21"
+last_updated = "2025-07-21"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = true
+license = "Apache 2.0"
+
+[limit]
+context = 262_144
+output = 81_920
+
+[modalities]
+input = ["text"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/Qwen/Qwen3-235B-A22B-Thinking-2507"
diff --git a/models/alibaba/qwen3-asr-flash.toml b/models/alibaba/qwen3-asr-flash.toml
new file mode 100644
index 00000000000..266d0372f58
--- /dev/null
+++ b/models/alibaba/qwen3-asr-flash.toml
@@ -0,0 +1,19 @@
+name = "Qwen3-ASR Flash"
+description = "Speech transcription model for accurate audio-to-text and captioning workflows"
+family = "qwen"
+release_date = "2025-09-08"
+last_updated = "2025-09-08"
+attachment = false
+reasoning = false
+temperature = false
+tool_call = false
+knowledge = "2024-04"
+open_weights = false
+
+[limit]
+context = 53_248
+output = 4_096
+
+[modalities]
+input = ["audio"]
+output = ["text"]
diff --git a/models/alibaba/qwen3-max-2026-01-23.toml b/models/alibaba/qwen3-max-2026-01-23.toml
new file mode 100644
index 00000000000..8f63001edd2
--- /dev/null
+++ b/models/alibaba/qwen3-max-2026-01-23.toml
@@ -0,0 +1,25 @@
+# https://docs.modelstudio.console.alibabacloud.com/en/model-studio/model-qwen3-max
+# https://www.alibabacloud.com/blog/pushing-qwen3-max-thinking-beyond-its-limits_602834
+name = "Qwen3 Max 2026-01-23"
+description = "Qwen3 Max snapshot integrating thinking and non-thinking modes for complex reasoning, instruction following, multilingual tasks, retrieval, and tool use"
+family = "qwen"
+release_date = "2026-01-23"
+last_updated = "2026-01-23"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+open_weights = false
+
+[limit]
+context = 262_144
+output = 65_536
+
+[modalities]
+input = ["text"]
+output = ["text"]
+
+[[links]]
+label = "Alibaba Cloud Model Studio"
+url = "https://docs.modelstudio.console.alibabacloud.com/en/model-studio/model-qwen3-max"
+type = "docs"
diff --git a/models/alibaba/qwen3-rerank.toml b/models/alibaba/qwen3-rerank.toml
new file mode 100644
index 00000000000..3c1da5d60ee
--- /dev/null
+++ b/models/alibaba/qwen3-rerank.toml
@@ -0,0 +1,17 @@
+name = "Qwen3 Rerank"
+description = "Qwen3 Rerank is a new proprietary model of the Qwen model family. Specifically designed for reranking tasks, built on the Qwen3 foundation model. Leveraging Qwen3’s robust multilingual text understanding capabilities, the series achieves state-of-the-art performance across multiple benchmarks for text embedding and reranking tasks."
+release_date = "2026-01-08"
+last_updated = "2026-01-08"
+attachment = false
+reasoning = false
+temperature = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 8_192
+output = 8_192
+
+[modalities]
+input = ["text"]
+output = ["text"]
diff --git a/models/alibaba/qwen3-vl-embedding.toml b/models/alibaba/qwen3-vl-embedding.toml
new file mode 100644
index 00000000000..a3048f96de5
--- /dev/null
+++ b/models/alibaba/qwen3-vl-embedding.toml
@@ -0,0 +1,17 @@
+name = "Qwen3 VL Embedding"
+description = "The Qwen3-VL-Embedding model series are the latest additions to the Qwen family, built upon the recently open-sourced and powerful Qwen3-VL foundation model. Specifically designed for multimodal information retrieval and cross-modal understanding, this suite accepts diverse inputs including text, images, screenshots, and videos, as well as inputs containing a mixture of these modalities."
+release_date = "2026-01-08"
+last_updated = "2026-01-08"
+attachment = true
+reasoning = false
+temperature = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 8_192
+output = 8_192
+
+[modalities]
+input = ["text", "video", "image"]
+output = ["text"]
diff --git a/models/alibaba/qwen3-vl-rerank.toml b/models/alibaba/qwen3-vl-rerank.toml
new file mode 100644
index 00000000000..c15ef8c3ab8
--- /dev/null
+++ b/models/alibaba/qwen3-vl-rerank.toml
@@ -0,0 +1,18 @@
+# https://docs.modelstudio.console.alibabacloud.com/en/model-studio/qwen3-vl-rerank
+name = "Qwen3 VL Rerank"
+description = "The Qwen3-VL-Rerank model series are the latest additions to the Qwen family, built upon the recently open-sourced and powerful Qwen3-VL foundation model. Specifically designed for multimodal information retrieval and cross-modal understanding, this suite accepts diverse inputs including text, images, screenshots, and videos, as well as inputs containing a mixture of these modalities."
+release_date = "2026-01-08"
+last_updated = "2026-01-08"
+attachment = true
+reasoning = false
+temperature = false
+tool_call = false
+open_weights = false
+
+[limit]
+context = 120_000
+output = 0
+
+[modalities]
+input = ["text", "image", "video"]
+output = ["text"]
diff --git a/providers/zenmux/models/qwen/qwen-audio-3.0-tts-plus.toml b/providers/zenmux/models/qwen/qwen-audio-3.0-tts-plus.toml
new file mode 100644
index 00000000000..d777ab2af16
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen-audio-3.0-tts-plus.toml
@@ -0,0 +1,3 @@
+# ZenMux pricing is non-token-denominated for this route and is intentionally omitted.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+base_model = "alibaba/qwen-audio-3.0-tts-plus"
diff --git a/providers/zenmux/models/qwen/qwen-image-2.0-pro.toml b/providers/zenmux/models/qwen/qwen-image-2.0-pro.toml
new file mode 100644
index 00000000000..b88491e9195
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen-image-2.0-pro.toml
@@ -0,0 +1,7 @@
+# ZenMux pricing is non-token-denominated for this route and is intentionally omitted.
+# Host prompt limit: the 2026-09-19 ZenMux page reports context_length=1024.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+base_model = "alibaba/qwen-image-2.0-pro"
+
+[limit]
+context = 1_024
diff --git a/providers/zenmux/models/qwen/qwen-image-2.0.toml b/providers/zenmux/models/qwen/qwen-image-2.0.toml
new file mode 100644
index 00000000000..5e791df4085
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen-image-2.0.toml
@@ -0,0 +1,7 @@
+# ZenMux pricing is non-token-denominated for this route and is intentionally omitted.
+# Host prompt limit: the 2026-09-19 ZenMux page reports context_length=1024.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+base_model = "alibaba/qwen-image-2.0"
+
+[limit]
+context = 1_024
diff --git a/providers/zenmux/models/qwen/qwen-image-3.0-pro.toml b/providers/zenmux/models/qwen/qwen-image-3.0-pro.toml
new file mode 100644
index 00000000000..ea890d3d706
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen-image-3.0-pro.toml
@@ -0,0 +1,7 @@
+# ZenMux pricing is non-token-denominated for this route and is intentionally omitted.
+# Host prompt limit: the 2026-09-19 ZenMux page reports context_length=1024.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+base_model = "alibaba/qwen-image-3.0-pro"
+
+[limit]
+context = 1_024
diff --git a/providers/zenmux/models/qwen/qwen-image-3.0.toml b/providers/zenmux/models/qwen/qwen-image-3.0.toml
new file mode 100644
index 00000000000..030aa99f9bb
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen-image-3.0.toml
@@ -0,0 +1,7 @@
+# ZenMux pricing is non-token-denominated for this route and is intentionally omitted.
+# Host prompt limit: the 2026-09-19 ZenMux page reports context_length=1024.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+base_model = "alibaba/qwen-image-3.0"
+
+[limit]
+context = 1_024
diff --git a/providers/zenmux/models/qwen/qwen3-14b.toml b/providers/zenmux/models/qwen/qwen3-14b.toml
new file mode 100644
index 00000000000..7691a66fcb7
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen3-14b.toml
@@ -0,0 +1,21 @@
+# Toggle: reasoning.enabled = true|false
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "alibaba/qwen3-14b"
+name = "Qwen3-14B"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.14
+output = 1.4
+cache_read = 0.028
+
+[limit]
+context = 32_000
+output = 32_000
diff --git a/providers/zenmux/models/qwen/qwen3-235b-a22b-2507.toml b/providers/zenmux/models/qwen/qwen3-235b-a22b-2507.toml
new file mode 100644
index 00000000000..bbff8c04049
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen3-235b-a22b-2507.toml
@@ -0,0 +1,14 @@
+# Host limits: the 2026-09-19 ZenMux page reports context_length=256000 and max_completion_tokens=128000.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+base_model = "alibaba/qwen3-235b-a22b-instruct-2507"
+name = "Qwen3 235B A22B Instruct 2507"
+structured_output = true
+
+[cost]
+input = 0.28
+output = 1.11
+cache_read = 0.056
+
+[limit]
+context = 256_000
+output = 128_000
diff --git a/providers/zenmux/models/qwen/qwen3-235b-a22b-thinking-2507.toml b/providers/zenmux/models/qwen/qwen3-235b-a22b-thinking-2507.toml
new file mode 100644
index 00000000000..17f14e62fbb
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen3-235b-a22b-thinking-2507.toml
@@ -0,0 +1,18 @@
+# Host limits: the 2026-09-19 ZenMux page reports context_length=256000 and max_completion_tokens=128000.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "alibaba/qwen3-235b-a22b-thinking-2507"
+structured_output = true
+reasoning_options = []
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.28
+output = 2.78
+cache_read = 0.056
+
+[limit]
+context = 256_000
+output = 128_000
diff --git a/providers/zenmux/models/qwen/qwen3-asr-flash.toml b/providers/zenmux/models/qwen/qwen3-asr-flash.toml
new file mode 100644
index 00000000000..6acef26a6a5
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen3-asr-flash.toml
@@ -0,0 +1,13 @@
+# ZenMux's list flags this route as reasoning-capable, but the transcription API
+# exposes neither a reasoning request control nor a reasoning response channel;
+# lab and first-party metadata both classify Qwen3 ASR Flash as non-reasoning.
+# Host limits: page max_context=1000000 and max_completion_tokens=65536;
+# OpenAI list independently reports context_length=1000000.
+# ZenMux pricing is non-token-denominated for this route and is intentionally omitted.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+base_model = "alibaba/qwen3-asr-flash"
+name = "Qwen3 ASR Flash"
+
+[limit]
+context = 1_000_000
+output = 65_536
diff --git a/providers/zenmux/models/qwen/qwen3-coder-plus.toml b/providers/zenmux/models/qwen/qwen3-coder-plus.toml
index 2b327c36d7b..86a58aadea9 100644
--- a/providers/zenmux/models/qwen/qwen3-coder-plus.toml
+++ b/providers/zenmux/models/qwen/qwen3-coder-plus.toml
@@ -1,24 +1,34 @@
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+base_model = "alibaba/qwen3-coder-plus"
 name = "Qwen3-Coder-Plus"
-description = "Qwen coding model for software agents, repository edits, and code reasoning"
-release_date = "2025-07-23"
-last_updated = "2025-07-23"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = false
+structured_output = true
 
 [cost]
-input = 1.00
-output = 5.00
-cache_read = 0.10
+input = 1
+output = 5
+cache_read = 0.1
 cache_write = 1.25
 
-[limit]
-context = 1000_000
-output = 64_000
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 1.8
+output = 9
+cache_read = 0.18
+cache_write = 2.25
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_000 }
+input = 3
+output = 15
+cache_read = 0.3
+cache_write = 3.75
 
-[modalities]
-input = ["text"]
-output = ["text"]
+[[cost.tiers]]
+tier = { type = "context", size = 256_000 }
+input = 6
+output = 60
+cache_read = 0.6
+cache_write = 7.5
+
+[limit]
+context = 1_000_000
diff --git a/providers/zenmux/models/qwen/qwen3-coder.toml b/providers/zenmux/models/qwen/qwen3-coder.toml
new file mode 100644
index 00000000000..661da6b6538
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen3-coder.toml
@@ -0,0 +1,14 @@
+# Host limits: the 2026-09-19 ZenMux page reports context_length=256000 and max_completion_tokens=128000.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+base_model = "alibaba/qwen3-coder-480b-a35b-instruct"
+name = "Qwen3-Coder"
+structured_output = true
+
+[cost]
+input = 1.25
+output = 5.01
+cache_read = 0.25
+
+[limit]
+context = 256_000
+output = 128_000
diff --git a/providers/zenmux/models/qwen/qwen3-max.toml b/providers/zenmux/models/qwen/qwen3-max.toml
index 4687663742c..5ebb5d782a0 100644
--- a/providers/zenmux/models/qwen/qwen3-max.toml
+++ b/providers/zenmux/models/qwen/qwen3-max.toml
@@ -1,23 +1,38 @@
+# Toggle: reasoning.enabled = true|false
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# Host limits: the 2026-09-19 ZenMux page reports context_length=256000 and max_completion_tokens=32000.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "alibaba/qwen3-max-2026-01-23"
 name = "Qwen3-Max-Thinking"
-description = "Qwen reasoning model for deliberate problem solving, math, and coding"
-release_date = "2026-01-23"
-last_updated = "2026-01-23"
-attachment = false
-reasoning = true
-reasoning_options = []
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = false
+structured_output = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
-input = 1.20
-output = 6.00
+input = 1.2
+output = 6
+cache_read = 0.12
+cache_write = 1.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 2.4
+output = 12
+cache_read = 0.24
+cache_write = 3
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_000 }
+input = 3
+output = 15
+cache_read = 0.3
+cache_write = 3.75
 
 [limit]
 context = 256_000
-output = 64_000
-
-[modalities]
-input = ["text"]
-output = ["text"]
+output = 32_000
diff --git a/providers/zenmux/models/qwen/qwen3-rerank.toml b/providers/zenmux/models/qwen/qwen3-rerank.toml
new file mode 100644
index 00000000000..01088e0fecd
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen3-rerank.toml
@@ -0,0 +1,6 @@
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+base_model = "alibaba/qwen3-rerank"
+
+[cost]
+input = 0.1
+output = 0
diff --git a/providers/zenmux/models/qwen/qwen3-vl-embedding.toml b/providers/zenmux/models/qwen/qwen3-vl-embedding.toml
new file mode 100644
index 00000000000..593f756a189
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen3-vl-embedding.toml
@@ -0,0 +1,6 @@
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+base_model = "alibaba/qwen3-vl-embedding"
+
+[cost]
+input = 0.1
+output = 0
diff --git a/providers/zenmux/models/qwen/qwen3-vl-plus.toml b/providers/zenmux/models/qwen/qwen3-vl-plus.toml
new file mode 100644
index 00000000000..84d1677542c
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen3-vl-plus.toml
@@ -0,0 +1,38 @@
+# Toggle: reasoning.enabled = true|false
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# Host modalities: the 2026-09-19 ZenMux page reports text, image, and video input.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "alibaba/qwen3-vl-plus"
+name = "Qwen3-VL-Plus"
+attachment = true
+structured_output = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.2
+output = 1.6
+cache_read = 0.02
+cache_write = 0.25
+
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 0.3
+output = 2.4
+cache_read = 0.03
+cache_write = 0.375
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_000 }
+input = 0.6
+output = 4.8
+cache_read = 0.06
+cache_write = 0.75
+
+[modalities]
+input = ["text", "image", "video"]
diff --git a/providers/zenmux/models/qwen/qwen3-vl-rerank.toml b/providers/zenmux/models/qwen/qwen3-vl-rerank.toml
new file mode 100644
index 00000000000..e88dfc4bd9a
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen3-vl-rerank.toml
@@ -0,0 +1,10 @@
+# Host limit: the 2026-09-19 ZenMux page reports context_length=1000000.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+base_model = "alibaba/qwen3-vl-rerank"
+
+[cost]
+input = 0.1
+output = 0
+
+[limit]
+context = 1_000_000
diff --git a/providers/zenmux/models/qwen/qwen3.5-flash.toml b/providers/zenmux/models/qwen/qwen3.5-flash.toml
index c863eed08b1..7babd4e43d7 100644
--- a/providers/zenmux/models/qwen/qwen3.5-flash.toml
+++ b/providers/zenmux/models/qwen/qwen3.5-flash.toml
@@ -1,22 +1,26 @@
-name = "Qwen3.5 Flash"
-description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
-release_date = "2026-03-20"
-last_updated = "2026-03-20"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = false
+# Toggle: reasoning.enabled = true|false
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# Host limits: the 2026-09-19 ZenMux page reports context_length=1024000 and max_completion_tokens=1024000.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "alibaba/qwen3.5-flash"
+name = "Qwen3.5-Flash"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
-input = 0.10
-output = 0.40
+input = 0.1
+output = 0.4
+cache_read = 0.01
+cache_write = 0.125
 
 [limit]
-context = 1_020_000
-output = 1_020_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
+context = 1_024_000
+output = 1_024_000
diff --git a/providers/zenmux/models/qwen/qwen3.5-plus.toml b/providers/zenmux/models/qwen/qwen3.5-plus.toml
index 0cdf87788f3..09c1e88bbf2 100644
--- a/providers/zenmux/models/qwen/qwen3.5-plus.toml
+++ b/providers/zenmux/models/qwen/qwen3.5-plus.toml
@@ -1,23 +1,30 @@
-name = "Qwen3.5 Plus"
-description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
-release_date = "2026-03-20"
-last_updated = "2026-03-20"
+# Toggle: reasoning.enabled = true|false
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "alibaba/qwen3.5-plus"
+name = "Qwen3.5-Plus"
 attachment = true
-reasoning = true
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high"] }]
-temperature = true
-tool_call = true
-knowledge = "2025-01-01"
-open_weights = false
+structured_output = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
-input = 0.80
-output = 4.80
+input = 0.4
+output = 2.4
+cache_read = 0.04
+cache_write = 0.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 256_000 }
+input = 0.5
+output = 3
+cache_read = 0.05
+cache_write = 0.625
 
 [limit]
-context = 1_000_000
 output = 64_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
diff --git a/providers/zenmux/models/qwen/qwen3.6-flash.toml b/providers/zenmux/models/qwen/qwen3.6-flash.toml
new file mode 100644
index 00000000000..b18da445fdb
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen3.6-flash.toml
@@ -0,0 +1,24 @@
+# Toggle: reasoning.enabled = true|false
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "alibaba/qwen3.6-flash"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.25
+output = 1.5
+cache_read = 0.025
+cache_write = 0.3125
+
+[[cost.tiers]]
+tier = { type = "context", size = 256_000 }
+input = 1
+output = 4
+cache_read = 0.1
+cache_write = 1.25
diff --git a/providers/zenmux/models/qwen/qwen3.6-max-preview.toml b/providers/zenmux/models/qwen/qwen3.6-max-preview.toml
new file mode 100644
index 00000000000..11ce923426e
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen3.6-max-preview.toml
@@ -0,0 +1,25 @@
+# Toggle: reasoning.enabled = true|false
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "alibaba/qwen3.6-max-preview"
+structured_output = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 1.3
+output = 7.8
+cache_read = 0.13
+cache_write = 1.625
+
+[[cost.tiers]]
+tier = { type = "context", size = 128_000 }
+input = 2
+output = 12
+cache_read = 0.2
+cache_write = 2.5
diff --git a/providers/zenmux/models/qwen/qwen3.6-plus.toml b/providers/zenmux/models/qwen/qwen3.6-plus.toml
index 6233134b5ac..525cfb1b421 100644
--- a/providers/zenmux/models/qwen/qwen3.6-plus.toml
+++ b/providers/zenmux/models/qwen/qwen3.6-plus.toml
@@ -1,31 +1,34 @@
+# Toggle: reasoning.enabled = true|false
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# ZenMux's 2026-09-19 page and both live OpenAI-/Anthropic-compatible lists
+# explicitly include file input; the catalog represents that host modality as PDF.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "alibaba/qwen3.6-plus"
 name = "Qwen3.6-Plus"
-description = "Qwen instruction model for multilingual chat, reasoning, and tool use"
-release_date = "2026-03-30"
-last_updated = "2026-03-30"
-attachment = false
-reasoning = true
-reasoning_options = [{ type = "toggle" }]
-temperature = true
-tool_call = true
-open_weights = false
+structured_output = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
-input = 0.50
-output = 3.00
+input = 0.5
+output = 3
 cache_read = 0.05
 cache_write = 0.625
 
 [[cost.tiers]]
-tier = { size = 256_000 }
-input = 2.00
-output = 6.00
-cache_read = 0.20
-cache_write = 2.50
+tier = { type = "context", size = 256_000 }
+input = 2
+output = 6
+cache_read = 0.2
+cache_write = 2.5
 
 [limit]
-context = 1_000_000
 output = 64_000
 
 [modalities]
-input = ["text"]
-output = ["text"]
+input = ["text", "image", "pdf", "video"]
diff --git a/providers/zenmux/models/qwen/qwen3.7-flash.toml b/providers/zenmux/models/qwen/qwen3.7-flash.toml
new file mode 100644
index 00000000000..6cc899e503a
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen3.7-flash.toml
@@ -0,0 +1,35 @@
+# Toggle: reasoning.enabled = true|false
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "alibaba/qwen3.7-flash"
+name = "Qwen3.7-Flash"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
+
+[cost]
+input = 0.03
+output = 0.13
+cache_read = 0.003
+cache_write = 0.038
+
+[[cost.tiers]]
+tier = { type = "context", size = 32_000 }
+input = 0.1
+output = 0.4
+cache_read = 0.01
+cache_write = 0.125
+
+[[cost.tiers]]
+tier = { type = "context", size = 256_000 }
+input = 0.2
+output = 0.8
+cache_read = 0.02
+cache_write = 0.25
+
+[limit]
+output = 64_000
diff --git a/providers/zenmux/models/qwen/qwen3.7-max.toml b/providers/zenmux/models/qwen/qwen3.7-max.toml
index 7a085d238f1..34e397e0ba7 100644
--- a/providers/zenmux/models/qwen/qwen3.7-max.toml
+++ b/providers/zenmux/models/qwen/qwen3.7-max.toml
@@ -1,8 +1,21 @@
+# Toggle: reasoning.enabled = true|false
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
 base_model = "alibaba/qwen3.7-max"
-reasoning_options = [{ type = "toggle" }]
+name = "Qwen3.7-Max"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
-input = 2.50
-output = 7.50
-cache_read = 0.50
-cache_write = 3.125
+input = 1.25
+output = 3.75
+cache_read = 0.125
+cache_write = 1.5625
+
+[limit]
+output = 64_000
diff --git a/providers/zenmux/models/qwen/qwen3.7-plus.toml b/providers/zenmux/models/qwen/qwen3.7-plus.toml
index 661f1915cc1..64bb828e576 100644
--- a/providers/zenmux/models/qwen/qwen3.7-plus.toml
+++ b/providers/zenmux/models/qwen/qwen3.7-plus.toml
@@ -1,15 +1,25 @@
+# Toggle: reasoning.enabled = true|false
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
 base_model = "alibaba/qwen3.7-plus"
-reasoning_options = [{ type = "toggle" }]
+name = "Qwen3.7-Plus"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
 
 [cost]
-input = 0.40
-output = 1.60
-cache_read = 0.08
-cache_write = 0.50
+input = 0.4
+output = 1.6
+cache_read = 0.04
+cache_write = 0.5
 
 [[cost.tiers]]
-tier = { size = 256_000 }
-input = 1.20
-output = 4.80
-cache_read = 0.24
-cache_write = 1.50
+tier = { type = "context", size = 256_000 }
+input = 1.2
+output = 4.8
+cache_read = 0.12
+cache_write = 1.5
diff --git a/providers/zenmux/models/qwen/qwen3.8-2.4t-a95b.toml b/providers/zenmux/models/qwen/qwen3.8-2.4t-a95b.toml
new file mode 100644
index 00000000000..fd60945a2cc
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen3.8-2.4t-a95b.toml
@@ -0,0 +1,19 @@
+# Effort: reasoning_effort = low|medium|xhigh
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "alibaba/qwen3.8-2.4t-a95b"
+name = "Qwen3.8-2.4T-A95B"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.17
+cache_write = 2.5
+
+[limit]
+context = 1_000_000
+output = 64_000
diff --git a/providers/zenmux/models/qwen/qwen3.8-27b.toml b/providers/zenmux/models/qwen/qwen3.8-27b.toml
new file mode 100644
index 00000000000..c574b301e05
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen3.8-27b.toml
@@ -0,0 +1,23 @@
+# Toggle: reasoning.enabled = true|false
+# Effort: reasoning_effort = low|medium|xhigh
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "alibaba/qwen3.8-27b"
+name = "Qwen3.8-27B"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[cost]
+input = 0.5
+output = 3
+cache_read = 0.05
+cache_write = 0.625
+
+[limit]
+context = 1_000_000
+output = 131_072
diff --git a/providers/zenmux/models/qwen/qwen3.8-flash.toml b/providers/zenmux/models/qwen/qwen3.8-flash.toml
new file mode 100644
index 00000000000..32dbb1e3a6e
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen3.8-flash.toml
@@ -0,0 +1,32 @@
+# Toggle: reasoning.enabled = true|false
+# Effort: reasoning_effort = low|medium|xhigh
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# ZenMux's 2026-09-19 page explicitly reports max_completion_tokens=1000000 for
+# this route; the 1M output limit is therefore an intentional host override.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "alibaba/qwen3.8-flash"
+name = "Qwen3.8-Flash"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+max = 262_144
+
+[cost]
+input = 0.16
+output = 0.47
+cache_read = 0.016
+cache_write = 0.2
+
+[limit]
+output = 1_000_000
diff --git a/providers/zenmux/models/qwen/qwen3.8-max-0902.toml b/providers/zenmux/models/qwen/qwen3.8-max-0902.toml
new file mode 100644
index 00000000000..4b710b76a1c
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen3.8-max-0902.toml
@@ -0,0 +1,33 @@
+# Toggle: reasoning.enabled = true|false
+# Effort: reasoning_effort = low|medium|xhigh
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# ZenMux's 2026-09-19 page and both live OpenAI-/Anthropic-compatible lists
+# advertise text, image, and video input only; PDF is intentionally omitted.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "alibaba/qwen3.8-max-0902"
+name = "Qwen3.8-Max-0902"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = 0
+max = 262_144
+
+[cost]
+input = 2
+output = 6
+cache_read = 0.17
+cache_write = 2.5
+
+[limit]
+output = 128_000
+
+[modalities]
+input = ["text", "image", "video"]
diff --git a/providers/zenmux/models/qwen/qwen3.8-max.toml b/providers/zenmux/models/qwen/qwen3.8-max.toml
new file mode 100644
index 00000000000..7c5111f7b93
--- /dev/null
+++ b/providers/zenmux/models/qwen/qwen3.8-max.toml
@@ -0,0 +1,36 @@
+# Toggle: reasoning.enabled = true|false
+# Effort: reasoning_effort = low|medium|xhigh
+# Budget: reasoning.max_tokens (integer reasoning tokens)
+# ZenMux's 2026-09-19 page and both live OpenAI-/Anthropic-compatible lists
+# advertise text, image, and video input only; PDF is intentionally omitted.
+# https://zenmux.ai/docs/api/openai/openai-list-models.html
+# https://zenmux.ai/docs/guide/advanced/reasoning.html
+base_model = "alibaba/qwen3.8-max"
+name = "Qwen3.8-Max"
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "xhigh"]
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = 0
+max = 262_144
+
+[cost]
+input = 1.4
+output = 4.2
+cache_read = 0.119
+cache_write = 1.75
+
+[limit]
+output = 64_000
+
+[modalities]
+input = ["text", "image", "video"]