chore(sync): update OpenRouter model catalog (#5316)

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

models/deepseek/deepseek-v4-pro-0813.toml

@@ -1,14 +1,16 @@
+# https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro-0813
 name = "DeepSeek V4 Pro 0813"
 description = "DeepSeek V4 Pro snapshot with million-token context and support for thinking and non-thinking modes"
 family = "deepseek-thinking"
 release_date = "2026-08-12"
-last_updated = "2026-08-12"
+last_updated = "2026-08-22"
 attachment = false
 reasoning = true
 temperature = true
 tool_call = true
 structured_output = true
-open_weights = false
+open_weights = true
+license = "MIT"
 
 [limit]
 context = 1_000_000
@@ -17,3 +19,7 @@ output = 384_000
 [modalities]
 input = ["text"]
 output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro-0813"

providers/crof/models/glm-4.7-flash.toml

@@ -1,16 +0,0 @@
-base_model = "zhipuai/glm-4.7-flash"
-# Crof's request docs establish no model-specific toggle, effort, or budget.
-# https://crof.ai/docs.md (accessed 2026-06-25)
-reasoning_options = []
-
-[cost]
-input = 0.04
-output = 0.3
-cache_read = 0.008
-cache_write = 0
-
-[limit]
-context = 202_752
-
-[provider]
-npm = "@ai-sdk/openai-compatible"

providers/crof/models/glm-4.7.toml

@@ -1,20 +0,0 @@
-base_model = "zhipuai/glm-4.7"
-# Crof's request docs establish no model-specific toggle, effort, or budget.
-# https://crof.ai/docs.md (accessed 2026-06-25)
-reasoning_options = []
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0.25
-output = 1.1
-cache_read = 0.05
-cache_write = 0
-
-[limit]
-context = 202_752
-output = 202_752
-
-[provider]
-npm = "@ai-sdk/openai-compatible"

providers/crof/models/glm-5.toml

@@ -1,20 +0,0 @@
-base_model = "zhipuai/glm-5"
-# Crof's request docs establish no model-specific toggle, effort, or budget.
-# https://crof.ai/docs.md (accessed 2026-06-25)
-reasoning_options = []
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0.48
-output = 1.9
-cache_read = 0.1
-cache_write = 0
-
-[limit]
-context = 202_752
-output = 202_752
-
-[provider]
-npm = "@ai-sdk/openai-compatible"

providers/crof/models/kimi-k2.5-lightning.toml

@@ -1,28 +0,0 @@
-name = "Kimi K2.5 (Lightning)"
-description = "Kimi multimodal agent model for visual understanding, coding, and planning"
-family = "kimi-k2"
-release_date = "2026-02-06"
-last_updated = "2026-02-06"
-attachment = true
-reasoning = true
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
-temperature = false
-tool_call = true
-structured_output = true
-open_weights = true
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 1.00
-output = 3.00
-cache_read = 0.20
-
-[limit]
-context = 131_072
-output = 32_768
-
-[modalities]
-input = ["text", "image", "video"]
-output = ["text"]

providers/crof/models/minimax-m2.5.toml

@@ -1,14 +0,0 @@
-base_model = "minimax/MiniMax-M2.5"
-# Crof's request docs establish no model-specific toggle, effort, or budget.
-# https://crof.ai/docs.md (accessed 2026-06-25)
-reasoning = true
-reasoning_options = []
-
-[cost]
-input = 0.11
-output = 0.95
-cache_read = 0.02
-cache_write = 0.375
-
-[provider]
-npm = "@ai-sdk/openai-compatible"

providers/crof/models/qwen3.8-27b.toml

@@ -1,13 +1,13 @@
-base_model = "moonshotai/kimi-k2.5"
+base_model = "alibaba/qwen3.8-27b"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
 
-[interleaved]
-field = "reasoning_content"
-
 [cost]
-input = 0.35
-output = 1.7
-cache_read = 0.07
+input = 0.25
+output = 2.1
+cache_read = 0.06
+
+[limit]
+output = 262_144
 
 [provider]
 npm = "@ai-sdk/openai-compatible"

providers/edenai/models/deepinfra/deepseek-ai/DeepSeek-V4-Pro-0813.toml

@@ -1,6 +1,4 @@
 base_model = "deepseek/deepseek-v4-pro-0813"
-tool_call = false
-structured_output = false
 
 [[reasoning_options]]
 type = "effort"

providers/edenai/models/qwen/deepseek-v4-flash-0731.toml

@@ -5,5 +5,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.176
-output = 0.528
+input = 0.352
+output = 1.056

providers/edenai/models/qwen/deepseek-v4-pro-0813.toml

@@ -5,5 +5,5 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.627
-output = 1.881
+input = 1.1616
+output = 3.4848

providers/kilo/models/google/gemma-4-26b-a4b-it.toml

@@ -6,8 +6,8 @@ type = "effort"
 values = ["none", "high"]
 
 [cost]
-input = 0.05
-output = 0.25
+input = 0.042
+output = 0.22
 
 [limit]
 output = 16_384

providers/kilo/models/qwen/qwen3-30b-a3b.toml

@@ -11,4 +11,4 @@ input = 0.13
 output = 0.52
 
 [limit]
-output = 8_192
+context = 40_960

providers/kilo/models/qwen/qwen3.5-122b-a10b.toml

@@ -9,5 +9,8 @@ values = ["none", "high"]
 input = 0.26
 output = 2.08
 
+[limit]
+output = 262_144
+
 [modalities]
 input = ["text", "image", "video"]

providers/kilo/models/tencent/hy-mt2-7b.toml

@@ -0,0 +1,23 @@
+name = "Tencent: Hy-MT2-7B"
+description = "Hy-MT2-7B is a 7B-parameter translation model from Tencent. It supports 33 language pairs and five Chinese dialect and minority-language pairs, with workflows for structured, delimiter-based, contextual, glossary-based, and style-guided translation."
+family = "Hy"
+release_date = "2026-08-19"
+last_updated = "2026-08-19"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = false
+structured_output = true
+open_weights = false
+
+[cost]
+input = 0.074
+output = 0.295
+
+[limit]
+context = 8_192
+output = 4_096
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/kilo/models/~deepseek/deepseek-v4-flash-latest.toml

@@ -15,13 +15,13 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.065
-output = 0.18
-cache_read = 0.02
+input = 0.04
+output = 0.13
+cache_read = 0.01
 
 [limit]
-context = 1_048_576
-output = 1_048_576
+context = 262_144
+output = 262_144
 
 [modalities]
 input = ["text"]

providers/llmgateway-providers/models/deepinfra/mimo-v2.5-pro.toml

@@ -0,0 +1,25 @@
+base_model = "xiaomi/mimo-v2.5-pro"
+name = "MiMo V2.5 Pro (DeepInfra)"
+tool_call = false
+structured_output = false
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 1
+output = 3
+cache_read = 0.2
+
+[[cost.tiers]]
+tier = { type = "context", size = 256_000 }
+input = 2
+output = 6
+cache_read = 0.4
+
+[limit]
+output = 16_384

providers/llmgateway-providers/models/deepinfra/mimo-v2.5.toml

@@ -0,0 +1,25 @@
+base_model = "xiaomi/mimo-v2.5"
+name = "MiMo V2.5 (DeepInfra)"
+structured_output = true
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
+
+[cost]
+input = 0.4
+output = 2
+cache_read = 0.08
+
+[[cost.tiers]]
+tier = { type = "context", size = 256_000 }
+input = 0.8
+output = 4
+cache_read = 0.16
+
+[limit]
+context = 262_144
+output = 16_384

providers/llmgateway-providers/models/novita/mimo-v2.5-pro.toml

@@ -0,0 +1,21 @@
+base_model = "xiaomi/mimo-v2.5-pro"
+name = "MiMo V2.5 Pro (NovitaAI)"
+structured_output = false
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "minimal", "low", "medium", "high"]
+
+[cost]
+input = 0.522
+output = 1.044
+cache_read = 0.0043
+
+[[cost.tiers]]
+tier = { type = "context", size = 256_000 }
+input = 2
+output = 6
+cache_read = 0.4

providers/llmgateway-providers/models/novita/mimo-v2.5.toml

@@ -0,0 +1,21 @@
+base_model = "xiaomi/mimo-v2.5"
+name = "MiMo V2.5 (NovitaAI)"
+structured_output = false
+
+[interleaved]
+field = "reasoning_content"
+
+[[reasoning_options]]
+type = "effort"
+values = ["none", "low", "medium", "high"]
+
+[cost]
+input = 0.168
+output = 0.336
+cache_read = 0.0034
+
+[[cost.tiers]]
+tier = { type = "context", size = 256_000 }
+input = 0.8
+output = 4
+cache_read = 0.16

providers/llmgateway-providers/models/openai/gpt-4o-mini-search-preview.toml

@@ -1,23 +0,0 @@
-name = "GPT-4o Mini Search Preview (OpenAI)"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "gpt"
-release_date = "2024-10-01"
-last_updated = "2024-10-01"
-attachment = true
-reasoning = false
-temperature = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.15
-output = 0.6
-
-[limit]
-context = 128_000
-output = 16_384
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

providers/llmgateway-providers/models/openai/gpt-4o-search-preview.toml

@@ -1,23 +0,0 @@
-name = "GPT-4o Search Preview (OpenAI)"
-description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
-family = "gpt"
-release_date = "2024-10-01"
-last_updated = "2024-10-01"
-attachment = true
-reasoning = false
-temperature = false
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 2.5
-output = 10
-
-[limit]
-context = 128_000
-output = 16_384
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

providers/llmgateway/models/gpt-4o-mini-search-preview.toml

@@ -1,23 +0,0 @@
-name = "GPT-4o Mini Search Preview"
-description = "Compact GPT model for low-latency assistance and high-volume workloads"
-family = "gpt"
-release_date = "2024-10-01"
-last_updated = "2024-10-01"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 0.15
-output = 0.60
-
-[limit]
-context = 128_000
-output = 16_384
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

providers/llmgateway/models/gpt-4o-search-preview.toml

@@ -1,23 +0,0 @@
-name = "GPT-4o Search Preview"
-description = "GPT model for general reasoning, writing, coding, and tool-assisted tasks"
-family = "gpt"
-release_date = "2024-10-01"
-last_updated = "2024-10-01"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = false
-structured_output = false
-open_weights = false
-
-[cost]
-input = 2.50
-output = 10.00
-
-[limit]
-context = 128_000
-output = 16_384
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

providers/llmgateway/models/mimo-v2.5-pro.toml

@@ -8,7 +8,7 @@ field = "reasoning_content"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "low", "medium", "high"]
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 0.435
@@ -20,6 +20,3 @@ tier = { type = "context", size = 256_000 }
 input = 2
 output = 6
 cache_read = 0.4
-
-[limit]
-context = 1_000_000

providers/llmgateway/models/mimo-v2.5.toml

@@ -8,7 +8,7 @@ field = "reasoning_content"
 
 [[reasoning_options]]
 type = "effort"
-values = ["none", "low", "medium", "high"]
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 0.14
@@ -20,6 +20,3 @@ tier = { type = "context", size = 256_000 }
 input = 0.8
 output = 4
 cache_read = 0.16
-
-[limit]
-context = 1_000_000

providers/nano-gpt/models/deepseek/deepseek-v4-flash-vision-exp.toml

@@ -5,9 +5,9 @@ type = "effort"
 values = ["none", "low", "high", "max"]
 
 [cost]
-input = 0.22
-output = 0.66
-cache_read = 0.007
+input = 0.44
+output = 1.32
+cache_read = 0.014
 
 [limit]
 context = 1_048_576

providers/nano-gpt/models/google/gemini-flash-latest.toml

@@ -1,22 +1,16 @@
 # Not included in subscription
 # Tracks Gemini 3.5 Flash (current flash-latest target).
 base_model = "google/gemini-flash-latest"
-tool_call = false
-structured_output = false
 
 [[reasoning_options]]
 type = "effort"
 values = ["minimal", "low", "medium", "high"]
 
 [cost]
-input = 0.75
-output = 3.75
-cache_read = 0.075
-cache_write = 0.041667
+input = 0.375
+output = 1.875
+cache_read = 0.0375
+cache_write = 0.020833
 
 [limit]
-context = 1_048_756
-input = 1_048_756
-
-[modalities]
-input = ["text", "image", "audio"]
+input = 1_048_576

providers/nano-gpt/models/google/gemma-4-26b-a4b-uncensored.toml

@@ -0,0 +1,24 @@
+name = "Gemma 4 26B A4B Uncensored"
+description = "Gemma 4 26B A4B Uncensored is an NVFP4 open-weight multimodal mixture-of-experts model tuned for fewer refusals across chat, coding, tool use, and long-context work."
+family = "gemma"
+release_date = "2026-08-22"
+last_updated = "2026-08-22"
+attachment = true
+reasoning = false
+tool_call = true
+structured_output = true
+open_weights = true
+
+[cost]
+input = 0.08
+output = 0.33
+cache_read = 0.04
+
+[limit]
+context = 131_072
+input = 131_072
+output = 32_768
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]

providers/nvidia/models/deepseek-ai/deepseek-v4-flash-0731.toml

@@ -0,0 +1,17 @@
+# NIM Chat schema: `reasoning_effort = none|high|max`; `none` disables thinking.
+# https://docs.api.nvidia.com/nim/reference/deepseek-ai-deepseek-v4-flash-infer
+#
+# Pricing: hosted on NVIDIA's API trial tier and currently free — no separate
+# list rate is published for this ID. Source: this model's catalog card,
+# https://build.nvidia.com/deepseek-ai/deepseek-v4-flash-0731 (governed by the
+# NVIDIA API Trial Terms of Service).
+base_model = "deepseek/deepseek-v4-flash-0731"
+
+reasoning_options = [{ type = "effort", values = ["none", "high", "max"] }]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.0
+output = 0.0

providers/nvidia/models/moonshotai/kimi-k3.toml

@@ -0,0 +1,16 @@
+# NIM wire syntax (verified against integrate.api.nvidia.com/v1, 2026-08-22):
+#   on/off: chat_template_kwargs = { "thinking": true | false }  (default: true)
+#   effort: top-level "reasoning_effort" = "low" | "high" | "max"
+base_model = "moonshotai/kimi-k3"
+
+reasoning_options = [
+  { type = "toggle" },
+  { type = "effort", values = ["low", "high", "max"] },
+]
+
+[interleaved]
+field = "reasoning_content"
+
+[cost]
+input = 0.0
+output = 0.0

providers/ofox/models/deepseek/deepseek-v4-flash-0731.toml

@@ -0,0 +1,17 @@
+# https://ofox.ai/models/deepseek/deepseek-v4-flash-0731
+# https://api-docs.deepseek.com/quick_start/pricing/
+# Toggle: thinking.type = enabled|disabled
+# Effort: reasoning_effort = low|high|max (Flash; lab maps xhigh→high)
+base_model = "deepseek/deepseek-v4-flash-0731"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "effort"
+values = ["low", "high", "max"]
+
+[cost]
+input = 0.44
+output = 1.32
+cache_read = 0.014

providers/openrouter/models/deepseek/deepseek-v4-flash.toml

@@ -11,9 +11,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.06006
-output = 0.12012
-cache_read = 0.012012
+input = 0.05166
+output = 0.10332
+cache_read = 0.010332
 
 [limit]
 context = 1_048_576

providers/openrouter/models/deepseek/deepseek-v4-pro-0813.toml

@@ -8,9 +8,9 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 1.188
-output = 3.564
-cache_read = 0.0396
+input = 1.122
+output = 3.366
+cache_read = 0.0374
 
 [limit]
 context = 1_048_576

providers/openrouter/models/deepseek/deepseek-v4-pro.toml

@@ -11,9 +11,9 @@ type = "effort"
 values = ["high", "xhigh"]
 
 [cost]
-input = 0.413772
-output = 0.827544
-cache_read = 0.034481
+input = 0.396894
+output = 0.793788
+cache_read = 0.033075
 
 [limit]
 context = 1_048_576

providers/openrouter/models/moonshotai/kimi-k2.6.toml

@@ -5,9 +5,9 @@ reasoning_options = []
 field = "reasoning_details"
 
 [cost]
-input = 0.5415
-output = 2.28
-cache_read = 0.0912
+input = 0.95
+output = 4
+cache_read = 0.16
 
 [modalities]
 input = ["text", "image"]

providers/openrouter/models/openai/gpt-oss-120b.toml

@@ -5,9 +5,8 @@ type = "effort"
 values = ["low", "medium", "high"]
 
 [cost]
-input = 0.03
+input = 0.037
 output = 0.17
-cache_read = 0.03
 
 [limit]
 output = 131_072

providers/openrouter/models/qwen/qwen3-30b-a3b.toml

@@ -6,8 +6,5 @@ structured_output = false
 type = "toggle"
 
 [cost]
-input = 0.13
-output = 0.52
-
-[limit]
-output = 8_192
+input = 0.12
+output = 0.5

providers/openrouter/models/qwen/qwen3.5-122b-a10b.toml

@@ -7,5 +7,8 @@ type = "toggle"
 input = 0.26
 output = 2.08
 
+[limit]
+output = 262_144
+
 [modalities]
 input = ["text", "image", "video"]

providers/openrouter/models/qwen/qwen3.8-27b.toml

@@ -9,8 +9,8 @@ type = "effort"
 values = ["low", "medium", "xhigh"]
 
 [cost]
-input = 0.45
-output = 3.2
+input = 0.4
+output = 3
 cache_read = 0.05
 
 [limit]

providers/openrouter/models/tencent/hy-mt2-7b.toml

@@ -0,0 +1,23 @@
+name = "Hy-MT2-7B"
+description = "Tencent Hy reasoning model for coding, instruction following, and agent tasks"
+family = "Hy"
+release_date = "2026-08-19"
+last_updated = "2026-08-19"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = false
+structured_output = true
+open_weights = true
+
+[cost]
+input = 0.074
+output = 0.295
+
+[limit]
+context = 8_192
+output = 4_096
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/openrouter/models/thinkingmachines/inkling.toml

@@ -7,9 +7,9 @@ type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "max"]
 
 [cost]
-input = 0.95
+input = 1
 output = 4.05
-cache_read = 0.16
+cache_read = 0.17
 
 [limit]
 output = 262_144

providers/openrouter/models/~deepseek/deepseek-v4-flash-latest.toml

@@ -18,13 +18,13 @@ type = "effort"
 values = ["low", "high", "max"]
 
 [cost]
-input = 0.065
-output = 0.18
-cache_read = 0.02
+input = 0.04
+output = 0.13
+cache_read = 0.01
 
 [limit]
 context = 1_310_720
-output = 1_048_576
+output = 262_144
 
 [modalities]
 input = ["text"]

providers/venice/models/deepseek-v4-pro-0813.toml

@@ -2,6 +2,7 @@ base_model = "deepseek/deepseek-v4-pro-0813"
 description = "Flagship DeepSeek model for coding, reasoning, and agentic work"
 release_date = "2026-08-14"
 last_updated = "2026-08-14"
+open_weights = false
 reasoning_options = []
 
 [cost]

providers/vercel/models/nvidia/nemotron-3.5-lightning-free.toml

@@ -0,0 +1,23 @@
+# Toggle: reasoning.enabled = true|false
+# Budget: reasoning.max_tokens (1..32768 reasoning tokens)
+# Sources:
+# - https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-chat-completions/reasoning
+# - https://docs.api.nvidia.com/nim/reference/nvidia-nemotron-3-5-lightning-30b-a3b-infer
+base_model = "nvidia/nemotron-3.5-lightning"
+name = "Nemotron 3.5 Lightning 30B (Free)"
+
+[[reasoning_options]]
+type = "toggle"
+
+[[reasoning_options]]
+type = "budget_tokens"
+min = 1
+max = 32_768
+
+[cost]
+input = 0
+output = 0
+
+[limit]
+context = 1_000_000
+output = 32_768

providers/vercel/models/nvidia/nemotron-3.5-lightning.toml

@@ -1,5 +1,8 @@
 # Toggle: reasoning.enabled = true|false
-# Source: https://build.nvidia.com/nvidia/nemotron-3.5-lightning-30b-a3b
+# Budget: reasoning.max_tokens (1..32768 reasoning tokens)
+# Sources:
+# - https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-chat-completions/reasoning
+# - https://docs.api.nvidia.com/nim/reference/nvidia-nemotron-3-5-lightning-30b-a3b-infer
 base_model = "nvidia/nemotron-3.5-lightning"
 name = "Nemotron 3.5 Lightning 30B"
 description = "Nemotron model for efficient reasoning, coding, and specialized AI agents"
@@ -7,10 +10,15 @@ description = "Nemotron model for efficient reasoning, coding, and specialized A
 [[reasoning_options]]
 type = "toggle"
 
+[[reasoning_options]]
+type = "budget_tokens"
+min = 1
+max = 32_768
+
 [cost]
-input = 0.05
-output = 0.2
-cache_read = 0.01
+input = 0
+output = 0
 
 [limit]
-output = 131_072
+context = 1_000_000
+output = 32_768