Add TopxAI provider (Claude, GPT, Grok, Kimi, GLM-5.3-Abliterated via one OpenAI-compatible endpoint)

providers/topxai/models/GLM-5.3-Abliterated.toml

@@ -0,0 +1,38 @@
+# Effort: $.reasoning_effort on /v1/chat/completions, forwarded unchanged; the base
+# GLM-5.3 always reasons with low|high|max. A commercially fine-tuned GLM-5.3 with its
+# refusal alignment removed, served from TopxAI's own private deployment rather than
+# Z.ai's API; text only. Unique to this host, so defined inline.
+# https://ai.topxea.com/docs/glm-5-3-abliterated (accessed 2026-09-19)
+name = "GLM-5.3-Abliterated"
+description = "Fine-tuned GLM-5.3 with refusal alignment removed, on TopxAI's private deployment; text only"
+family = "glm"
+release_date = "2026-09-18"
+last_updated = "2026-09-18"
+attachment = false
+reasoning = true
+temperature = true
+tool_call = true
+structured_output = true
+open_weights = false
+
+# The base GLM-5.3 always reasons with effort low|high|max; the fine-tune keeps
+# the same request shape on /v1/chat/completions ($.reasoning_effort).
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 4
+output = 7
+cache_read = 0.4
+
+[limit]
+context = 1_000_000
+output = 131_072
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/topxai/models/claude-fable-5-1.toml

@@ -0,0 +1,16 @@
+# Effort only, as the lab and peers author this model: $.output_config.effort on
+# /v1/messages, $.reasoning_effort on /v1/chat/completions, both forwarded unchanged.
+# https://ai.topxea.com/docs/api-reference (accessed 2026-09-19)
+# https://ai.topxea.com/pricing/claude-fable-5-1 (accessed 2026-09-19)
+base_model = "anthropic/claude-fable-5-1"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 5
+output = 25
+cache_read = 0.125
+cache_write = 6.25

providers/topxai/models/claude-fable-5.toml

@@ -0,0 +1,16 @@
+# Effort only, as the lab and peers author this model: $.output_config.effort on
+# /v1/messages, $.reasoning_effort on /v1/chat/completions, both forwarded unchanged.
+# https://ai.topxea.com/docs/api-reference (accessed 2026-09-19)
+# https://ai.topxea.com/pricing/claude-fable-5 (accessed 2026-09-19)
+base_model = "anthropic/claude-fable-5"
+structured_output = true
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25

providers/topxai/models/claude-opus-5.toml

@@ -0,0 +1,20 @@
+# Toggle: $.thinking.type = "enabled"|"disabled" on /v1/messages (the Anthropic surface
+# TopxAI relays byte for byte); effort: $.output_config.effort. On the registered
+# /v1/chat/completions surface the relay forwards $.reasoning_effort unchanged.
+# Wire fields and pass-through: https://ai.topxea.com/docs/api-reference (accessed 2026-09-19)
+# https://ai.topxea.com/pricing/claude-opus-5 (accessed 2026-09-19)
+base_model = "anthropic/claude-opus-5"
+structured_output = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2.5
+output = 12.5
+cache_read = 0.25
+cache_write = 3.125

providers/topxai/models/claude-sonnet-5.toml

@@ -0,0 +1,21 @@
+# Toggle: $.thinking.type = "enabled"|"disabled" on /v1/messages (the Anthropic surface
+# TopxAI relays byte for byte); effort: $.output_config.effort. On the registered
+# /v1/chat/completions surface the relay forwards $.reasoning_effort unchanged.
+# Wire fields and pass-through: https://ai.topxea.com/docs/api-reference (accessed 2026-09-19)
+# Shared-pool price, 50% of Anthropic's list; cache lanes from the same table.
+# https://ai.topxea.com/pricing/claude-sonnet-5 (accessed 2026-09-19)
+base_model = "anthropic/claude-sonnet-5"
+structured_output = true
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 1
+output = 5
+cache_read = 0.1
+cache_write = 1.25

providers/topxai/models/gpt-5.6-sol.toml

@@ -0,0 +1,23 @@
+# Effort: $.reasoning.effort on /v1/responses, $.reasoning_effort on /v1/chat/completions,
+# forwarded unchanged; the lab's level set for this model, including "none".
+# Shared-pool price, 50% of OpenAI's list. Requests whose input reaches 272,001
+# tokens bill the whole request at the tier below.
+# https://ai.topxea.com/pricing/gpt-5.6-sol (accessed 2026-09-19)
+base_model = "openai/gpt-5.6-sol"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 2
+output = 10
+cache_read = 0.2
+cache_write = 2.5
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 4
+output = 15
+cache_read = 0.4
+cache_write = 5

providers/topxai/models/gpt-6-astra.toml

@@ -0,0 +1,21 @@
+# Effort: $.reasoning.effort on /v1/responses, $.reasoning_effort on /v1/chat/completions,
+# forwarded unchanged; the lab's and peers' level set for this model.
+# https://ai.topxea.com/pricing/gpt-6-astra (accessed 2026-09-19)
+base_model = "openai/gpt-6-astra"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 5
+output = 25
+cache_read = 0.5
+cache_write = 6.25
+
+[[cost.tiers]]
+tier = { type = "context", size = 272_000 }
+input = 10
+output = 37.5
+cache_read = 1
+cache_write = 12.5

providers/topxai/models/grok-4.6.toml

@@ -0,0 +1,22 @@
+# Effort: $.reasoning.effort on /v1/responses, $.reasoning_effort on /v1/chat/completions,
+# forwarded unchanged. Shared-pool price, 50% of xAI's list. Requests whose input
+# reaches 200,000 tokens bill the whole request at the tier below.
+# https://ai.topxea.com/pricing/grok-4.6 (accessed 2026-09-19)
+base_model = "xai/grok-4.6"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "medium", "high", "xhigh"]
+
+[cost]
+input = 1
+output = 3
+cache_read = 0.25
+cache_write = 1
+
+[[cost.tiers]]
+tier = { type = "context", size = 200_000 }
+input = 2
+output = 6
+cache_read = 0.5
+cache_write = 2

providers/topxai/models/kimi-k3.toml

@@ -0,0 +1,20 @@
+# Toggle: $.thinking.type = "enabled"|"disabled"|"adaptive" as Moonshot's API takes it,
+# forwarded unchanged on /v1/chat/completions; adaptive depth via $.output_config.effort.
+# One official line at 80% of Moonshot's list.
+# https://ai.topxea.com/pricing/kimi-k3 (accessed 2026-09-19)
+base_model = "moonshotai/kimi-k3"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 2.4
+output = 12
+cache_read = 0.24

providers/topxai/provider.toml

@@ -0,0 +1,14 @@
+# TopxAI relays the labs' own models through one OpenAI- and Anthropic-compatible
+# endpoint at fixed USD prices on prepaid credit. Prices below are the Auto-route
+# price (the shared pool, 50% of the lab's list price); the official line is 90% of
+# list. Live table and JSON: https://ai.topxea.com/pricing and
+# https://ai.topxea.com/api/pricing (accessed 2026-09-19).
+# Raw Chat (/v1/chat/completions): $.reasoning_effort forwarded unchanged to the lab.
+# Raw Responses (/v1/responses, GPT and Grok): $.reasoning.effort forwarded unchanged.
+# Raw Messages (/v1/messages, Claude): $.thinking and $.output_config.effort forwarded unchanged.
+# https://ai.topxea.com/docs/api-reference (accessed 2026-09-19)
+name = "TopxAI"
+env = ["TOPXAI_API_KEY"]
+npm = "@ai-sdk/openai-compatible"
+api = "https://ai.topxea.com/v1"
+doc = "https://ai.topxea.com/docs"