diff --git a/data/models/aion-3.5-mini.json b/data/models/aion-3.5-mini.json
new file mode 100644
index 00000000..feb23c31
--- /dev/null
+++ b/data/models/aion-3.5-mini.json
@@ -0,0 +1,73 @@
+{
+  "id": "aion-3.5-mini",
+  "object": "model",
+  "created": 1790121600,
+  "owned_by": "aionlabs",
+  "name": "Aion 3.5 Mini",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": []
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "1M_tokens",
+    "input": 0.7,
+    "output": 1.4,
+    "cache_read": 0.18,
+    "cache_write": null
+  },
+  "features": [
+    "caching"
+  ],
+  "info": {
+    "status": "active",
+    "release_date": "2026-09-23",
+    "knowledge_cutoff": null,
+    "description": "Aion Labs API model released on 2026-09-23, with reasoning. The models page lists a 256K context, a 32K maximum output, and prices of $0.70 per 1M input tokens, $0.18 per 1M cached input tokens, and $1.40 per 1M output tokens. Aion Labs is not in the abbreviation table, so those limits are not converted. The API id is aion-labs/aion-3.5-mini. The documented host is https://api.aionlabs.ai; the page shows the models list and does not print a chat path or a cache-write price.",
+    "docs": "https://www.aionlabs.ai/docs/models/",
+    "verified_at": "2026-10-27"
+  },
+  "aliases": [
+    "aion-labs/aion-3.5-mini"
+  ],
+  "translations": {
+    "zh": {
+      "description": "Aion Labs API 模型，2026-09-23 发布，支持推理。模型页列出 256K 上下文、32K 最大输出，价格为输入每百万 tokens 0.70 美元、缓存输入 0.18 美元、输出 1.40 美元。Aion Labs 不在缩写换算表中，因此不换算这两项上限。API id 为 aion-labs/aion-3.5-mini。文档给出的主机是 https://api.aionlabs.ai；页面展示了模型列表，没有写出聊天路径或缓存写入价。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Models | Aion Labs",
+      "url": "https://www.aionlabs.ai/docs/models/",
+      "retrieved_at": "2026-10-27"
+    }
+  ],
+  "providers": [
+    {
+      "id": "aionlabs",
+      "model": "aion-labs/aion-3.5-mini",
+      "base_url": "https://api.aionlabs.ai",
+      "endpoints": {},
+      "reasoning_param": null
+    }
+  ]
+}
diff --git a/data/models/aion-3.5.json b/data/models/aion-3.5.json
new file mode 100644
index 00000000..d29fb88b
--- /dev/null
+++ b/data/models/aion-3.5.json
@@ -0,0 +1,73 @@
+{
+  "id": "aion-3.5",
+  "object": "model",
+  "created": 1790121600,
+  "owned_by": "aionlabs",
+  "name": "Aion 3.5",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": []
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "1M_tokens",
+    "input": 3,
+    "output": 6,
+    "cache_read": 0.75,
+    "cache_write": null
+  },
+  "features": [
+    "caching"
+  ],
+  "info": {
+    "status": "active",
+    "release_date": "2026-09-23",
+    "knowledge_cutoff": null,
+    "description": "Aion Labs API model released on 2026-09-23, with reasoning. The models page lists a 256K context, a 32K maximum output, and prices of $3.00 per 1M input tokens, $0.75 per 1M cached input tokens, and $6.00 per 1M output tokens. Aion Labs is not in the abbreviation table, so those limits are not converted. The API id is aion-labs/aion-3.5. The documented host is https://api.aionlabs.ai; the page shows the models list and does not print a chat path or a cache-write price.",
+    "docs": "https://www.aionlabs.ai/docs/models/",
+    "verified_at": "2026-10-27"
+  },
+  "aliases": [
+    "aion-labs/aion-3.5"
+  ],
+  "translations": {
+    "zh": {
+      "description": "Aion Labs API 模型，2026-09-23 发布，支持推理。模型页列出 256K 上下文、32K 最大输出，价格为输入每百万 tokens 3.00 美元、缓存输入 0.75 美元、输出 6.00 美元。Aion Labs 不在缩写换算表中，因此不换算这两项上限。API id 为 aion-labs/aion-3.5。文档给出的主机是 https://api.aionlabs.ai；页面展示了模型列表，没有写出聊天路径或缓存写入价。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Models | Aion Labs",
+      "url": "https://www.aionlabs.ai/docs/models/",
+      "retrieved_at": "2026-10-27"
+    }
+  ],
+  "providers": [
+    {
+      "id": "aionlabs",
+      "model": "aion-labs/aion-3.5",
+      "base_url": "https://api.aionlabs.ai",
+      "endpoints": {},
+      "reasoning_param": null
+    }
+  ]
+}
diff --git a/data/models/beam-1.0.json b/data/models/beam-1.0.json
new file mode 100644
index 00000000..d6a5fce8
--- /dev/null
+++ b/data/models/beam-1.0.json
@@ -0,0 +1,74 @@
+{
+  "id": "beam-1.0",
+  "object": "model",
+  "created": 1792108800,
+  "owned_by": "reflection",
+  "name": "Beam 1.0",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": [
+      "low",
+      "medium",
+      "high",
+      "xhigh",
+      "max"
+    ]
+  },
+  "pricing": null,
+  "features": [
+    "tools"
+  ],
+  "info": {
+    "status": "active",
+    "release_date": "2026-10-16",
+    "knowledge_cutoff": "2026-06-30",
+    "description": "Open-weight sparse mixture-of-experts model with 501.3B total parameters and 23.4B active parameters. It takes text input and generates text, with a 1M-token context length and a knowledge cutoff of 2026-06-30. Weights are Apache 2.0. Reasoning values are low, medium (default), high, xhigh, and max, and the model supports tool calling. No first-party API price is published.",
+    "docs": "https://huggingface.co/reflection/Beam-1.0-501B-A23B",
+    "verified_at": "2026-10-22"
+  },
+  "aliases": [
+    "Beam-1.0-501B-A23B"
+  ],
+  "translations": {
+    "zh": {
+      "description": "开源稀疏混合专家模型，模型卡写明总参数 501.3B、激活参数 23.4B，输入和输出均为文本，上下文长度 1M tokens，知识截止日期 2026-06-30。权重为 Apache 2.0。推理档位为 low、medium（默认）、high、xhigh 和 max，并支持工具调用。未公布第一方 API 价格。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Introducing Beam | Reflection",
+      "url": "https://reflection.ai/blog/introducing-beam",
+      "retrieved_at": "2026-10-22"
+    },
+    {
+      "title": "Beam 1.0 model card | Hugging Face",
+      "url": "https://huggingface.co/reflection/Beam-1.0-501B-A23B",
+      "retrieved_at": "2026-10-22"
+    },
+    {
+      "title": "Introducing Beam | Hugging Face Blog",
+      "url": "https://huggingface.co/blog/reflection-beam",
+      "retrieved_at": "2026-10-22"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/models/carbon-x.json b/data/models/carbon-x.json
new file mode 100644
index 00000000..07d0d002
--- /dev/null
+++ b/data/models/carbon-x.json
@@ -0,0 +1,54 @@
+{
+  "id": "carbon-x",
+  "object": "model",
+  "created": null,
+  "owned_by": "huggingface-bio",
+  "name": "Carbon-X",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": 40960,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": []
+  },
+  "pricing": null,
+  "features": [],
+  "info": {
+    "status": "active",
+    "release_date": null,
+    "knowledge_cutoff": null,
+    "description": "Hugging Face Bio Carbon-X is a 14.8B decoder-only model trained on DNA and natural-language text, with a 40,960-token context window, under the Apache 2.0 license. Input and output are text. The chat template supports enable_thinking, and the model was post-trained and evaluated with thinking disabled. No first-party API price or calendar release date is published.",
+    "docs": "https://huggingface.co/HuggingFaceBio/Carbon-X",
+    "verified_at": "2026-10-24"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "Hugging Face Bio 的 Carbon-X 是 14.8B 解码器模型，在 DNA 与自然语言文本上联合训练，上下文窗口 40,960 tokens，许可证为 Apache 2.0。输入和输出都是文本。对话模板支持 enable_thinking，后训练和评测时关闭思考。官方未公布第一方 API 价格，也没有日历发布日期。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Carbon-X model card | Hugging Face Bio",
+      "url": "https://huggingface.co/HuggingFaceBio/Carbon-X",
+      "retrieved_at": "2026-10-24"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/models/deepseek-flash.json b/data/models/deepseek-flash.json
index a13e1d97..08477b84 100644
--- a/data/models/deepseek-flash.json
+++ b/data/models/deepseek-flash.json
@@ -3,7 +3,7 @@
   "object": "model",
   "created": 1788998400,
   "owned_by": "deepseek",
-  "name": "DeepSeek-V4.1-Flash",
+  "name": "DeepSeek-V4.1-Flash-1017",
   "api": {
     "types": [
       "chat",
@@ -59,9 +59,9 @@
     "status": "active",
     "release_date": "2026-09-10",
     "knowledge_cutoff": null,
-    "description": "DeepSeek-V4.1-Flash, the smallest model of the new architecture family, with native multimodal visual understanding.",
+    "description": "As of 2026-10-17, deepseek-flash is served by DeepSeek-V4.1-Flash-1017. The API name and peak pricing are unchanged. The pricing page lists a 1M-token context, a 384K maximum output, native vision, JSON output, and tool calls. Weights are released under the MIT License. Off-peak rates are half the peak rates stored here.",
     "docs": "https://api-docs.deepseek.com/quick_start/pricing",
-    "verified_at": "2026-10-11"
+    "verified_at": "2026-10-22"
   },
   "aliases": [
     "deepseek-v4-flash",
@@ -69,19 +69,24 @@
   ],
   "translations": {
     "zh": {
-      "description": "DeepSeek-V4.1-Flash，新架构系列中最小的模型，具备原生多模态视觉理解能力。"
+      "description": "自 2026-10-17 起，deepseek-flash 由 DeepSeek-V4.1-Flash-1017 提供。API 名称和高峰价格不变。定价页写明上下文 1M tokens、最大输出 384K，并支持原生视觉、JSON 输出和工具调用。权重以 MIT 许可发布。非高峰价格是此处所记高峰价的一半。"
     }
   },
   "sources": [
     {
       "title": "Models & Pricing | DeepSeek API Docs",
       "url": "https://api-docs.deepseek.com/quick_start/pricing",
-      "retrieved_at": "2026-10-10"
+      "retrieved_at": "2026-10-22"
     },
     {
       "title": "Change Log | DeepSeek API Docs",
       "url": "https://api-docs.deepseek.com/updates",
-      "retrieved_at": "2026-10-10"
+      "retrieved_at": "2026-10-22"
+    },
+    {
+      "title": "DeepSeek-V4.1-Flash-1017 model card (MIT License) | Hugging Face",
+      "url": "https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017",
+      "retrieved_at": "2026-10-22"
     },
     {
       "title": "Thinking Mode | DeepSeek API Docs",
diff --git a/data/models/ember-1.json b/data/models/ember-1.json
new file mode 100644
index 00000000..5ab5ba42
--- /dev/null
+++ b/data/models/ember-1.json
@@ -0,0 +1,82 @@
+{
+  "id": "ember-1",
+  "object": "model",
+  "created": 1790035200,
+  "owned_by": "fireworks",
+  "name": "Ember-1",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "image"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": null,
+    "efforts": []
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "1M_tokens",
+    "input": 3,
+    "output": 15,
+    "cache_read": 0.3,
+    "cache_write": null
+  },
+  "features": [
+    "tools",
+    "caching"
+  ],
+  "info": {
+    "status": "active",
+    "release_date": "2026-09-22",
+    "knowledge_cutoff": null,
+    "description": "Serverless base model. The model page says it was created on 22 September 2026 and is Ready. Serverless pricing is $3.00 per 1M input tokens, $0.30 per 1M cached input tokens, and $15.00 per 1M output tokens. Image input and function calling are supported. The page lists the context length as 1040k tokens. Fireworks is not in the abbreviation table and no exact integer is printed, so the context limit is unset. No cache-write price or maximum output is published.",
+    "docs": "https://fireworks.ai/models/fireworks/ember-1",
+    "verified_at": "2026-10-27"
+  },
+  "aliases": [
+    "fireworks/ember-1"
+  ],
+  "translations": {
+    "zh": {
+      "description": "无服务器基础模型。模型页写明创建于 2026 年 9 月 22 日，状态为 Ready。无服务器价格为输入每百万 tokens 3.00 美元、缓存输入 0.30 美元、输出 15.00 美元。支持图像输入和函数调用。页面把上下文长度写成 1040k tokens。Fireworks 不在缩写换算表中，页面也没有精确整数，因此上下文上限留空。没有公布缓存写入价或最大输出。"
+    }
+  },
+  "sources": [
+    {
+      "title": "fireworks/ember-1 | Fireworks",
+      "url": "https://fireworks.ai/models/fireworks/ember-1",
+      "retrieved_at": "2026-10-27"
+    },
+    {
+      "title": "Querying text models | Fireworks Docs",
+      "url": "https://docs.fireworks.ai/guides/querying-text-models",
+      "retrieved_at": "2026-10-27"
+    }
+  ],
+  "providers": [
+    {
+      "id": "fireworks",
+      "model": "fireworks/ember-1",
+      "base_url": "https://api.fireworks.ai/inference/v1",
+      "endpoints": {
+        "chat": "POST /chat/completions"
+      },
+      "reasoning_param": null
+    }
+  ]
+}
diff --git a/data/models/flint.json b/data/models/flint.json
new file mode 100644
index 00000000..0d579794
--- /dev/null
+++ b/data/models/flint.json
@@ -0,0 +1,54 @@
+{
+  "id": "flint",
+  "object": "model",
+  "created": 1792540800,
+  "owned_by": "recursive",
+  "name": "Recursive Flint",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": null,
+    "efforts": []
+  },
+  "pricing": null,
+  "features": [],
+  "info": {
+    "status": "active",
+    "release_date": "2026-10-21",
+    "knowledge_cutoff": null,
+    "description": "Open-weight mixture-of-experts coding model with 40 billion total parameters and 2.5 billion active per token, under the Apache 2.0 license. The announcement states a context window of up to 512k tokens and recommends staying within 256k tokens. No first-party API or price is published.",
+    "docs": "https://recursive.com/updates/introducing-recursive-flint",
+    "verified_at": "2026-10-22"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "开源混合专家编程模型，公告写明总参数 400 亿、每 token 激活 25 亿，许可为 Apache 2.0。上下文窗口最高 512k tokens，并建议编程和其他任务保持在 256k tokens 以内。未公布第一方 API 或价格。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Introducing Flint | Recursive",
+      "url": "https://recursive.com/updates/introducing-recursive-flint",
+      "retrieved_at": "2026-10-22"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/models/flux-3-image-klein.json b/data/models/flux-3-image-klein.json
new file mode 100644
index 00000000..8a07185a
--- /dev/null
+++ b/data/models/flux-3-image-klein.json
@@ -0,0 +1,87 @@
+{
+  "id": "flux-3-image-klein",
+  "object": "model",
+  "created": 1792022400,
+  "owned_by": "black-forest-labs",
+  "name": "FLUX 3 Image Klein",
+  "api": {
+    "types": [
+      "images"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "image"
+    ],
+    "output": [
+      "image"
+    ]
+  },
+  "reasoning": {
+    "supported": null,
+    "efforts": []
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "image",
+    "input": null,
+    "output": 0.008,
+    "cache_read": null,
+    "cache_write": null
+  },
+  "features": [],
+  "info": {
+    "status": "active",
+    "release_date": "2026-10-15",
+    "knowledge_cutoff": null,
+    "description": "Open-weight 5.6B FLUX 3 image model for text-to-image, editing, and up to 10 reference images, under the FLUX Kommunity License. The API model id is flux-3-image-klein. Official per-image prices by output size are $0.008 at 768×768, $0.009 at 1K, $0.011 at 1.5K, $0.014 at 2K, and $0.036 at 4K. The stored price is the lowest tier, 768×768.",
+    "docs": "https://docs.bfl.ml/flux_3/flux3_image_klein",
+    "verified_at": "2026-10-16"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "开放权重的 5.6B FLUX 3 图像模型，支持文生图、编辑和最多 10 张参考图，许可证为 FLUX Kommunity License。API 模型 ID 为 flux-3-image-klein。官方按输出尺寸的单张价格为 768×768 $0.008、1K $0.009、1.5K $0.011、2K $0.014、4K $0.036。记录的是最低档 768×768。"
+    }
+  },
+  "sources": [
+    {
+      "title": "FLUX 3 Image Klein",
+      "url": "https://bfl.ai/blog/flux-3-image-klein",
+      "retrieved_at": "2026-10-16"
+    },
+    {
+      "title": "Pricing（FLUX 3 Klein：768×768 $0.008、1K $0.009、1.5K $0.011、2K $0.014、4K $0.036）| Black Forest Labs",
+      "url": "https://docs.bfl.ml/quick_start/pricing",
+      "retrieved_at": "2026-10-16"
+    },
+    {
+      "title": "FLUX 3 Image Klein | Black Forest Labs Docs",
+      "url": "https://docs.bfl.ml/flux_3/flux3_image_klein",
+      "retrieved_at": "2026-10-16"
+    },
+    {
+      "title": "Generate an image with FLUX 3 Klein | Black Forest Labs API",
+      "url": "https://docs.bfl.ml/api-reference/utility/generate-an-image-with-flux-3-klein",
+      "retrieved_at": "2026-10-16"
+    }
+  ],
+  "providers": [
+    {
+      "id": "black-forest-labs",
+      "model": "flux-3-image-klein",
+      "base_url": "https://api.bfl.ai",
+      "endpoints": {
+        "images": "POST /v1/flux-3-image-klein"
+      },
+      "reasoning_param": null
+    }
+  ]
+}
diff --git a/data/models/gemini-3.8-flash-cyber.json b/data/models/gemini-3.8-flash-cyber.json
new file mode 100644
index 00000000..f4ebb4ac
--- /dev/null
+++ b/data/models/gemini-3.8-flash-cyber.json
@@ -0,0 +1,73 @@
+{
+  "id": "gemini-3.8-flash-cyber",
+  "object": "model",
+  "created": 1789516800,
+  "owned_by": "google",
+  "name": "Gemini 3.8 Flash Cyber",
+  "api": {
+    "types": [
+      "generate",
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": 1048576,
+    "input": null,
+    "output": 65536
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "image",
+      "audio",
+      "video"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": []
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "1M_tokens",
+    "input": 1.5,
+    "output": 7.5,
+    "cache_read": 0.15,
+    "cache_write": null
+  },
+  "features": [
+    "structured_output",
+    "caching"
+  ],
+  "info": {
+    "status": "active",
+    "release_date": "2026-09-16",
+    "knowledge_cutoff": null,
+    "description": "Generally available to allowlisted projects. Input is text, image, audio, and video, and output is text. The context window is 1,048,576 tokens and the maximum output is 65,536 tokens. Thinking, structured output, context caching, GenerateContent, and Chat Completions are supported. Function calling, grounding, code execution, URL context, computer use, batch inference, the Interactions API, and the Live API are not supported. The recorded price is the Agent Platform Global standard rate, which is the same at or below 200K input tokens and above 200K: $1.50 per 1M input tokens, $0.15 per 1M cached input tokens, and $7.50 per 1M output tokens. Non-global standard rates are $1.65, $0.165, and $8.25. Priority and Flex/Batch rates are not recorded, and the page does not publish a cache-write price or a request URL.",
+    "docs": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/gemini/3-8-flash-cyber",
+    "verified_at": "2026-10-27"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "对白名单项目正式可用。输入为文本、图像、音频和视频，输出为文本。上下文窗口 1,048,576 tokens，最大输出 65,536 tokens。支持思考、结构化输出、上下文缓存、GenerateContent 和 Chat Completions。不支持函数调用、Grounding、代码执行、URL context、计算机操作、批量推理、Interactions API 和 Live API。记录的是 Agent Platform Global 标准价，200K 输入 tokens 及以下与以上相同：输入每百万 tokens 1.50 美元，缓存输入 0.15 美元，输出 7.50 美元。非 Global 标准价为 1.65、0.165 和 8.25 美元。Priority 与 Flex/Batch 价格不记录。页面没有给出缓存写入价或请求 URL。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Gemini 3.8 Flash Cyber | Gemini Enterprise Agent Platform",
+      "url": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/gemini/3-8-flash-cyber",
+      "retrieved_at": "2026-10-27"
+    },
+    {
+      "title": "Agent Platform Pricing | Google Cloud",
+      "url": "https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing",
+      "retrieved_at": "2026-10-27"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/models/gemini-3.8-flash.json b/data/models/gemini-3.8-flash.json
index fef7ae43..4b9d4986 100644
--- a/data/models/gemini-3.8-flash.json
+++ b/data/models/gemini-3.8-flash.json
@@ -43,9 +43,9 @@
   "pricing": {
     "currency": "USD",
     "unit": "1M_tokens",
-    "input": 0.75,
-    "output": 3.75,
-    "cache_read": 0.075,
+    "input": 1.5,
+    "output": 7.5,
+    "cache_read": 0.15,
     "cache_write": null
   },
   "features": [
@@ -59,14 +59,14 @@
     "status": "active",
     "release_date": "2026-09-02",
     "knowledge_cutoff": "2026-03",
-    "description": "Most intelligent Flash model for long-horizon coding, agents and enterprise workflows.",
+    "description": "Most intelligent Flash model for long-horizon coding, agents and enterprise workflows. Standard price is $1.50 input, $7.50 output, and $0.15 cache read per 1M tokens from January 1, 2027. The introductory price through December 31, 2026 is $0.75, $3.75, and $0.075.",
     "docs": "https://ai.google.dev/gemini-api/docs/models/gemini-3.8-flash",
-    "verified_at": "2026-10-11"
+    "verified_at": "2026-10-26"
   },
   "aliases": [],
   "translations": {
     "zh": {
-      "description": "最智能的 Flash 模型，适用于长周期编程、智能体和企业工作流。"
+      "description": "最智能的 Flash 模型，适用于长周期编程、智能体和企业工作流。2027 年 1 月 1 日起的标准价为每百万 token 输入 1.50 美元、输出 7.50 美元、缓存读取 0.15 美元。2026 年 12 月 31 日前的介绍价为 0.75、3.75 和 0.075 美元。"
     }
   },
   "sources": [
@@ -83,7 +83,12 @@
     {
       "title": "Pricing | Gemini API",
       "url": "https://ai.google.dev/gemini-api/docs/pricing",
-      "retrieved_at": "2026-10-10"
+      "retrieved_at": "2026-10-26"
+    },
+    {
+      "title": "What's new in Gemini 3.8 Flash | Gemini API",
+      "url": "https://ai.google.dev/gemini-api/docs/generate-content/whats-new-gemini-3.8",
+      "retrieved_at": "2026-10-26"
     },
     {
       "title": "Deprecations | Gemini API",
diff --git a/data/models/glimpse.json b/data/models/glimpse.json
new file mode 100644
index 00000000..d7dd9eca
--- /dev/null
+++ b/data/models/glimpse.json
@@ -0,0 +1,65 @@
+{
+  "id": "glimpse",
+  "object": "model",
+  "created": null,
+  "owned_by": "elorian",
+  "name": "Glimpse",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "image",
+      "video"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": []
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "1M_tokens",
+    "input": 0.6,
+    "output": 3,
+    "cache_read": null,
+    "cache_write": null
+  },
+  "features": [
+    "tools"
+  ],
+  "info": {
+    "status": "active",
+    "release_date": null,
+    "knowledge_cutoff": null,
+    "description": "A visual thinking model available through an API and a chat interface. It takes text, images, and video and reasons in visual space, including with multimodal tools. Input tokens cost $0.60 per million and output tokens $3.00 per million; the comparison table labels those rates uncached. The page does not publish a calendar date, a cached-token price, a context length, a maximum output, or a request URL.",
+    "docs": "https://elorian.ai/blog/introducing-glimpse",
+    "verified_at": "2026-10-27"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "可通过 API 和聊天界面使用的视觉思考模型。它接收文本、图像和视频，并在视觉空间中推理，包括使用多模态工具。输入每百万 tokens 0.60 美元，输出 3.00 美元；对照表把这两项标为未缓存价格。页面没有给出日历日期、缓存价、上下文长度、最大输出或请求 URL。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Introducing Glimpse | Elorian",
+      "url": "https://elorian.ai/blog/introducing-glimpse",
+      "retrieved_at": "2026-10-27"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/models/gpt-5.6-cyber.json b/data/models/gpt-5.6-cyber.json
index 15669ecc..49555cad 100644
--- a/data/models/gpt-5.6-cyber.json
+++ b/data/models/gpt-5.6-cyber.json
@@ -37,7 +37,7 @@
     "input": 12.5,
     "output": 75,
     "cache_read": 1.25,
-    "cache_write": null
+    "cache_write": 15.625
   },
   "features": [
     "tools",
@@ -51,7 +51,7 @@
     "knowledge_cutoff": "2026-02-16",
     "description": "Advanced cybersecurity model for authorized vulnerability research and security testing (approved access only).",
     "docs": "https://developers.openai.com/api/docs/models/gpt-5.6-cyber",
-    "verified_at": "2026-10-10"
+    "verified_at": "2026-10-16"
   },
   "aliases": [],
   "translations": {
@@ -66,9 +66,9 @@
       "retrieved_at": "2026-10-10"
     },
     {
-      "title": "Pricing | OpenAI API",
+      "title": "Pricing（gpt-5.6-cyber 短上下文缓存写入 $15.625／百万 tokens）| OpenAI API",
       "url": "https://developers.openai.com/api/docs/pricing",
-      "retrieved_at": "2026-10-10"
+      "retrieved_at": "2026-10-16"
     },
     {
       "title": "Models | OpenAI API",
diff --git a/data/models/gpt-6.1-terra.json b/data/models/gpt-6.1-terra.json
new file mode 100644
index 00000000..88fa8cbd
--- /dev/null
+++ b/data/models/gpt-6.1-terra.json
@@ -0,0 +1,107 @@
+{
+  "id": "gpt-6.1-terra",
+  "object": "model",
+  "created": 1792972800,
+  "owned_by": "openai",
+  "name": "GPT-6.1 Terra",
+  "api": {
+    "types": [
+      "chat",
+      "responses",
+      "batch"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": 1050000,
+    "input": 922000,
+    "output": 128000
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "image"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": [
+      "none",
+      "low",
+      "medium",
+      "high",
+      "xhigh",
+      "max"
+    ]
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "1M_tokens",
+    "input": 1,
+    "output": 5,
+    "cache_read": 0.05,
+    "cache_write": 1.25
+  },
+  "features": [
+    "tools",
+    "structured_output",
+    "streaming",
+    "caching",
+    "batch"
+  ],
+  "info": {
+    "status": "active",
+    "release_date": "2026-10-26",
+    "knowledge_cutoff": "2026-06-01",
+    "description": "A fast, efficient model for visual reasoning, precise computer use, and clear, natural writing. reasoning.effort accepts none, low, medium (default), high, xhigh, or max. Tool calling requires the Responses API. The recorded price is the standard short-context rate: $1 per 1M input tokens, $0.05 per 1M cached input tokens, $1.25 per 1M cache-write tokens for 30-minute retention, and $5 per 1M output tokens. Standard 12-hour cache writes are $2.50 per 1M tokens. When input exceeds 272K tokens, the full request uses the long-context standard rates: $2 input, $7.50 output, $0.10 cached input, $2.50 for a 30-minute cache write, and $5 for a 12-hour cache write. Batch and Flex are 50% lower, regional processing adds 10%, and fast mode is 2× standard; those rates are not recorded.",
+    "docs": "https://developers.openai.com/api/docs/models/gpt-6.1-terra",
+    "verified_at": "2026-10-27"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "面向视觉推理、精确计算机操作和自然写作的快速模型。reasoning.effort 可取 none、low、medium（默认）、high、xhigh 或 max。工具调用需要 Responses API。记录的是标准短上下文价格：输入每百万 tokens 1 美元，缓存输入 0.05 美元，30 分钟缓存写入 1.25 美元，输出 5 美元。标准 12 小时缓存写入为每百万 tokens 2.50 美元。输入超过 272K tokens 时，整次请求按长上下文标准价计费：输入 2 美元、输出 7.50 美元、缓存输入 0.10 美元、30 分钟缓存写入 2.50 美元、12 小时缓存写入 5 美元。Batch 与 Flex 低 50%，区域处理加 10%，快速模式为标准价的 2 倍；这些价格不记录。"
+    }
+  },
+  "sources": [
+    {
+      "title": "GPT-6.1 Terra | OpenAI API",
+      "url": "https://developers.openai.com/api/docs/models/gpt-6.1-terra",
+      "retrieved_at": "2026-10-27"
+    },
+    {
+      "title": "Pricing | OpenAI API",
+      "url": "https://developers.openai.com/api/docs/pricing",
+      "retrieved_at": "2026-10-27"
+    },
+    {
+      "title": "Introducing GPT-6.1 Terra | OpenAI",
+      "url": "https://openai.com/index/introducing-gpt-6-1-terra",
+      "retrieved_at": "2026-10-27"
+    },
+    {
+      "title": "Addendum to GPT-6 Astra System Card: GPT-6.1 Terra | OpenAI Deployment Safety Hub",
+      "url": "https://deploymentsafety.openai.com/gpt-6-1-terra",
+      "retrieved_at": "2026-10-27"
+    }
+  ],
+  "providers": [
+    {
+      "id": "openai",
+      "model": "gpt-6.1-terra",
+      "base_url": "https://api.openai.com",
+      "endpoints": {
+        "chat": "POST /v1/chat/completions",
+        "responses": "POST /v1/responses",
+        "batch": "POST /v1/batch"
+      },
+      "reasoning_param": "reasoning.effort",
+      "reasoning_params": {
+        "responses": "reasoning.effort"
+      }
+    }
+  ]
+}
diff --git a/data/models/hy3.json b/data/models/hy3.json
index ae378308..482fb547 100644
--- a/data/models/hy3.json
+++ b/data/models/hy3.json
@@ -54,9 +54,11 @@
     "knowledge_cutoff": null,
     "description": "Tencent Hy3 295B-A21B MoE model with switchable reasoning depth, served on TokenHub.",
     "docs": "https://cloud.tencent.com/document/product/1823/132252",
-    "verified_at": "2026-10-11"
+    "verified_at": "2026-10-26"
   },
-  "aliases": [],
+  "aliases": [
+    "hy3-preview"
+  ],
   "translations": {
     "zh": {
       "description": "腾讯 Hy3 295B-A21B MoE 模型，推理深度可切换，部署于 TokenHub。"
@@ -71,12 +73,17 @@
     {
       "title": "模型列表 | 腾讯云 TokenHub",
       "url": "https://cloud.tencent.com/document/product/1823/130051",
-      "retrieved_at": "2026-10-10"
+      "retrieved_at": "2026-10-26"
+    },
+    {
+      "title": "Hy Token Plan 中 Hy3 preview 自动路由到 Hy3 | 腾讯云公告",
+      "url": "https://cloud.tencent.com/announce/detail/2384",
+      "retrieved_at": "2026-10-26"
     },
     {
       "title": "模型定价 | 腾讯云 TokenHub",
       "url": "https://cloud.tencent.com/document/product/1823/130055",
-      "retrieved_at": "2026-10-10"
+      "retrieved_at": "2026-10-26"
     },
     {
       "title": "腾讯混元Hy3正式上线发布 | 腾讯云开发者社区",
diff --git a/data/models/kimi-k3.1-pro-highspeed.json b/data/models/kimi-k3.1-pro-highspeed.json
new file mode 100644
index 00000000..81277dc3
--- /dev/null
+++ b/data/models/kimi-k3.1-pro-highspeed.json
@@ -0,0 +1,119 @@
+{
+  "id": "kimi-k3.1-pro-highspeed",
+  "object": "model",
+  "created": null,
+  "owned_by": "moonshot",
+  "name": "Kimi K3.1 Pro Highspeed",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": 1048576,
+    "input": null,
+    "output": 1048576
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "image",
+      "video"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": [
+      "low",
+      "high",
+      "max"
+    ]
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "1M_tokens",
+    "input": 4,
+    "output": 20,
+    "cache_read": 0.2,
+    "cache_write": 4
+  },
+  "features": [
+    "tools",
+    "structured_output",
+    "caching",
+    "streaming"
+  ],
+  "info": {
+    "status": "active",
+    "release_date": null,
+    "knowledge_cutoff": null,
+    "description": "High-speed variant of Kimi K3.1 Pro that delivers 2-3× the speed of K3.1 Pro. It uses the same 1,048,576-token context, image and video input, and always-on thinking.",
+    "docs": "https://platform.kimi.ai/docs/guide/kimi-k3-1-pro-quickstart",
+    "verified_at": "2026-10-26"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "Kimi K3.1 Pro 的高速版本，速度约为 K3.1 Pro 的 2–3 倍。上下文同样为 1,048,576 tokens，支持图像和视频输入，思考模式始终开启。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Kimi K3.1 Pro Quickstart | Kimi API Platform",
+      "url": "https://platform.kimi.ai/docs/guide/kimi-k3-1-pro-quickstart",
+      "retrieved_at": "2026-10-26"
+    },
+    {
+      "title": "Kimi K3.1 Pro 快速开始 | Kimi API 开放平台",
+      "url": "https://platform.kimi.com/docs/guide/kimi-k3-1-pro-quickstart",
+      "retrieved_at": "2026-10-26"
+    },
+    {
+      "title": "Chat Model Pricing | Kimi API Platform",
+      "url": "https://platform.kimi.ai/docs/pricing/chat",
+      "retrieved_at": "2026-10-26"
+    },
+    {
+      "title": "对话模型定价 | Kimi API 开放平台",
+      "url": "https://platform.kimi.com/docs/pricing/chat",
+      "retrieved_at": "2026-10-26"
+    },
+    {
+      "title": "Introducing Kimi K3.1 Pro | Kimi",
+      "url": "https://www.kimi.ai/blog/kimi-k3-1-pro",
+      "retrieved_at": "2026-10-26"
+    }
+  ],
+  "providers": [
+    {
+      "id": "moonshot",
+      "model": "kimi-k3.1-pro-highspeed",
+      "base_url": "https://api.moonshot.ai",
+      "endpoints": {
+        "chat": "POST /v1/chat/completions"
+      },
+      "reasoning_param": "reasoning_effort"
+    },
+    {
+      "id": "moonshot",
+      "model": "kimi-k3.1-pro-highspeed",
+      "base_url": "https://api.moonshot.cn",
+      "endpoints": {
+        "chat": "POST /v1/chat/completions"
+      },
+      "reasoning_param": "reasoning_effort",
+      "pricing": {
+        "currency": "CNY",
+        "unit": "1M_tokens",
+        "input": 26,
+        "output": 136,
+        "cache_read": 1.3,
+        "cache_write": 26
+      }
+    }
+  ]
+}
diff --git a/data/models/kimi-k3.1-pro.json b/data/models/kimi-k3.1-pro.json
new file mode 100644
index 00000000..1da76e6f
--- /dev/null
+++ b/data/models/kimi-k3.1-pro.json
@@ -0,0 +1,119 @@
+{
+  "id": "kimi-k3.1-pro",
+  "object": "model",
+  "created": null,
+  "owned_by": "moonshot",
+  "name": "Kimi K3.1 Pro",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": 1048576,
+    "input": null,
+    "output": 1048576
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "image",
+      "video"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": [
+      "low",
+      "high",
+      "max"
+    ]
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "1M_tokens",
+    "input": 2,
+    "output": 10,
+    "cache_read": 0.2,
+    "cache_write": 2
+  },
+  "features": [
+    "tools",
+    "structured_output",
+    "caching",
+    "streaming"
+  ],
+  "info": {
+    "status": "active",
+    "release_date": null,
+    "knowledge_cutoff": null,
+    "description": "First model in the Kimi K3.1 series and Moonshot's most capable 3T-class open-weight model to date, with a 1,048,576-token context window. It improves coding, 3D creation, financial research, and other agentic work. Thinking is always on.",
+    "docs": "https://platform.kimi.ai/docs/guide/kimi-k3-1-pro-quickstart",
+    "verified_at": "2026-10-26"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "K3.1 系列的第一个模型，也是 Moonshot 迄今能力最强的 3 万亿级开源权重模型，上下文 1,048,576 tokens。它在代码、3D 创作、金融研究和其他长程智能体能力上有所提升。思考模式始终开启。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Kimi K3.1 Pro Quickstart | Kimi API Platform",
+      "url": "https://platform.kimi.ai/docs/guide/kimi-k3-1-pro-quickstart",
+      "retrieved_at": "2026-10-26"
+    },
+    {
+      "title": "Kimi K3.1 Pro 快速开始 | Kimi API 开放平台",
+      "url": "https://platform.kimi.com/docs/guide/kimi-k3-1-pro-quickstart",
+      "retrieved_at": "2026-10-26"
+    },
+    {
+      "title": "Chat Model Pricing | Kimi API Platform",
+      "url": "https://platform.kimi.ai/docs/pricing/chat",
+      "retrieved_at": "2026-10-26"
+    },
+    {
+      "title": "对话模型定价 | Kimi API 开放平台",
+      "url": "https://platform.kimi.com/docs/pricing/chat",
+      "retrieved_at": "2026-10-26"
+    },
+    {
+      "title": "Introducing Kimi K3.1 Pro | Kimi",
+      "url": "https://www.kimi.ai/blog/kimi-k3-1-pro",
+      "retrieved_at": "2026-10-26"
+    }
+  ],
+  "providers": [
+    {
+      "id": "moonshot",
+      "model": "kimi-k3.1-pro",
+      "base_url": "https://api.moonshot.ai",
+      "endpoints": {
+        "chat": "POST /v1/chat/completions"
+      },
+      "reasoning_param": "reasoning_effort"
+    },
+    {
+      "id": "moonshot",
+      "model": "kimi-k3.1-pro",
+      "base_url": "https://api.moonshot.cn",
+      "endpoints": {
+        "chat": "POST /v1/chat/completions"
+      },
+      "reasoning_param": "reasoning_effort",
+      "pricing": {
+        "currency": "CNY",
+        "unit": "1M_tokens",
+        "input": 13,
+        "output": 68,
+        "cache_read": 1.3,
+        "cache_write": 13
+      }
+    }
+  ]
+}
diff --git a/data/models/kolibri-1.json b/data/models/kolibri-1.json
new file mode 100644
index 00000000..54c1ef34
--- /dev/null
+++ b/data/models/kolibri-1.json
@@ -0,0 +1,64 @@
+{
+  "id": "kolibri-1",
+  "object": "model",
+  "created": 1790985600,
+  "owned_by": "aleph-alpha",
+  "name": "Kolibri-1",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": 1048576,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": [
+      "none",
+      "low",
+      "medium",
+      "high"
+    ]
+  },
+  "pricing": null,
+  "features": [
+    "tools",
+    "structured_output"
+  ],
+  "info": {
+    "status": "active",
+    "release_date": "2026-10-03",
+    "knowledge_cutoff": "2026-06-18",
+    "description": "Open-weight mixture-of-experts reasoning model for German and English, released on 3 October 2026 under Apache 2.0. The model card lists a context length of 1,048,576 tokens and recommends at most 262,144 tokens for serving efficiency. reasoning_effort accepts none, low, medium, or high, and the card documents tool calling and structured output. No standard API token price or first-party request URL is published.",
+    "docs": "https://huggingface.co/Aleph-Alpha/Kolibri-1",
+    "verified_at": "2026-10-27"
+  },
+  "aliases": [
+    "Aleph-Alpha/Kolibri-1"
+  ],
+  "translations": {
+    "zh": {
+      "description": "面向德语和英语的开放权重混合专家推理模型，2026 年 10 月 3 日以 Apache 2.0 发布。模型卡列出的上下文长度为 1,048,576 tokens，并建议为了服务效率最多使用 262,144 tokens。reasoning_effort 可取 none、low、medium 或 high，模型卡写明支持工具调用和结构化输出。没有公布 API 标准 token 价格或第一方请求 URL。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Aleph-Alpha/Kolibri-1 | Hugging Face",
+      "url": "https://huggingface.co/Aleph-Alpha/Kolibri-1",
+      "retrieved_at": "2026-10-27"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/models/lfm2.5-24b-a2b.json b/data/models/lfm2.5-24b-a2b.json
new file mode 100644
index 00000000..100a2465
--- /dev/null
+++ b/data/models/lfm2.5-24b-a2b.json
@@ -0,0 +1,71 @@
+{
+  "id": "lfm2.5-24b-a2b",
+  "object": "model",
+  "created": 1791849600,
+  "owned_by": "liquid",
+  "name": "LFM2.5-24B-A2B",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": 131072,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": []
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "1M_tokens",
+    "input": 0.1,
+    "output": 1,
+    "cache_read": 0.05,
+    "cache_write": null
+  },
+  "features": [
+    "tools",
+    "caching"
+  ],
+  "info": {
+    "status": "active",
+    "release_date": "2026-10-13",
+    "knowledge_cutoff": null,
+    "description": "Open-weight text model with 24B total parameters, about 2B active, and a 131,072-token context window. The model card documents tool use, and the model always thinks. Liquid Console API inference is priced at $0.10 per 1M input tokens, $0.05 per 1M cached input tokens, and $1.00 per 1M output tokens. A $2 signup credit is separate from that token price. The pages do not publish a cache-write price, a maximum output, or a request URL.",
+    "docs": "https://www.liquid.ai/blog/lfm2-5-24b-a2b",
+    "verified_at": "2026-10-27"
+  },
+  "aliases": [
+    "LiquidAI/LFM2.5-24B-A2B"
+  ],
+  "translations": {
+    "zh": {
+      "description": "开放权重文本模型，总参数 24B、激活约 2B，上下文窗口 131,072 tokens。模型卡写明支持工具使用，并且模型始终思考。Liquid Console API 推理价格为输入每百万 tokens 0.10 美元、缓存输入 0.05 美元、输出 1.00 美元。注册赠送的 2 美元额度与该 token 价格分开。页面没有给出缓存写入价、最大输出或请求 URL。"
+    }
+  },
+  "sources": [
+    {
+      "title": "LFM2.5-24B-A2B | Liquid AI",
+      "url": "https://www.liquid.ai/blog/lfm2-5-24b-a2b",
+      "retrieved_at": "2026-10-27"
+    },
+    {
+      "title": "LiquidAI/LFM2.5-24B-A2B | Hugging Face",
+      "url": "https://huggingface.co/LiquidAI/LFM2.5-24B-A2B",
+      "retrieved_at": "2026-10-27"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/models/mercury-3.json b/data/models/mercury-3.json
new file mode 100644
index 00000000..c45fb795
--- /dev/null
+++ b/data/models/mercury-3.json
@@ -0,0 +1,87 @@
+{
+  "id": "mercury-3",
+  "object": "model",
+  "created": 1792368000,
+  "owned_by": "inception",
+  "name": "Mercury 3",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": [
+      "inceptionai"
+    ]
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": 131072
+  },
+  "modalities": {
+    "input": [
+      "text"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": [
+      "instant",
+      "low",
+      "medium",
+      "high"
+    ]
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "1M_tokens",
+    "input": 0.2,
+    "output": 0.7,
+    "cache_read": 0.02,
+    "cache_write": null
+  },
+  "features": [
+    "tools",
+    "structured_output",
+    "caching"
+  ],
+  "info": {
+    "status": "active",
+    "release_date": "2026-10-19",
+    "knowledge_cutoff": null,
+    "description": "Diffusion language model with a 1M-token context window and a maximum of 131,072 output tokens. Reasoning effort accepts instant, low, medium, and high. Text in, text out, with tool calling and structured outputs. The recorded price is the standard rate; launch pricing is 50% lower.",
+    "docs": "https://docs.inceptionlabs.ai/get-started/models",
+    "verified_at": "2026-10-22"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "扩散语言模型，上下文窗口为 1M tokens，最大输出 131,072 tokens。reasoning effort 取值为 instant、low、medium 和 high。输入和输出均为文本，支持工具调用和结构化输出。此处记录的是标准价；首发价为其五折。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Models | Inception Docs",
+      "url": "https://docs.inceptionlabs.ai/get-started/models",
+      "retrieved_at": "2026-10-22"
+    },
+    {
+      "title": "Diffusion at Scale: Introducing Mercury 3 | Inception",
+      "url": "https://www.inceptionlabs.ai/blog/introducing-mercury-3",
+      "retrieved_at": "2026-10-22"
+    }
+  ],
+  "providers": [
+    {
+      "id": "inception",
+      "model": "mercury-3",
+      "base_url": "https://api.inceptionlabs.ai/v1",
+      "endpoints": {
+        "chat": "POST /chat/completions"
+      },
+      "reasoning_param": "reasoning_effort"
+    }
+  ]
+}
diff --git a/data/models/ming-3.0-asr-tiny.json b/data/models/ming-3.0-asr-tiny.json
new file mode 100644
index 00000000..4b212964
--- /dev/null
+++ b/data/models/ming-3.0-asr-tiny.json
@@ -0,0 +1,57 @@
+{
+  "id": "ming-3.0-asr-tiny",
+  "object": "model",
+  "created": null,
+  "owned_by": "inclusionai",
+  "name": "Ming-3.0-asr-tiny",
+  "api": {
+    "types": [
+      "audio"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "audio"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": null,
+    "efforts": []
+  },
+  "pricing": null,
+  "features": [
+    "streaming"
+  ],
+  "info": {
+    "status": "active",
+    "release_date": null,
+    "knowledge_cutoff": null,
+    "description": "inclusionAI speech-recognition model that pairs a Ling-tiny-v3 language backbone with an audio encoder. It transcribes Chinese and English, including dialects, and can take hotwords or dialogue context as text. Output is transcription text. The card says image and video inputs are unsupported, and that a self-hosted server can stream the transcript. Weights and original project code use the MIT license. Parameter count, context limit, release date, and a first-party hosted price are not stated.",
+    "docs": "https://huggingface.co/inclusionAI/Ming-3.0-asr-tiny",
+    "verified_at": "2026-10-24"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "inclusionAI 的语音识别模型，把 Ling-tiny-v3 语言骨干和音频编码器结合在一起。它转写中文和英文，包括方言，并可以把热词或对话上下文作为文本输入。输出是转写文本。模型卡写明不支持图像和视频输入，自托管服务可以流式返回转写。权重和项目原代码使用 MIT 许可证。参数量、上下文上限、发布日期和第一方托管价格均未写明。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Ming-3.0-asr-tiny model card | inclusionAI",
+      "url": "https://huggingface.co/inclusionAI/Ming-3.0-asr-tiny",
+      "retrieved_at": "2026-10-24"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/models/ming-3.0-audio-flash.json b/data/models/ming-3.0-audio-flash.json
new file mode 100644
index 00000000..79930619
--- /dev/null
+++ b/data/models/ming-3.0-audio-flash.json
@@ -0,0 +1,57 @@
+{
+  "id": "ming-3.0-audio-flash",
+  "object": "model",
+  "created": null,
+  "owned_by": "inclusionai",
+  "name": "Ming-3.0-audio-flash",
+  "api": {
+    "types": [
+      "chat",
+      "audio"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "audio"
+    ],
+    "output": [
+      "text",
+      "audio"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": []
+  },
+  "pricing": null,
+  "features": [],
+  "info": {
+    "status": "active",
+    "release_date": null,
+    "knowledge_cutoff": null,
+    "description": "inclusionAI native audio-language mixture-of-experts model with 126B total parameters and 6.5B activated per step. It takes audio and text and can return text or text with synthesized speech. The card states audio reasoning and the MIT license. The 128K figure is a serving profile, not a published context limit. No release date or first-party hosted price is on the card.",
+    "docs": "https://huggingface.co/inclusionAI/Ming-3.0-audio-flash",
+    "verified_at": "2026-10-24"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "inclusionAI 的原生音频语言模型，混合专家架构，总参数 126B，每步激活 6.5B。输入为音频和文本，可以返回文本，或文本加合成语音。模型卡写明支持音频推理，许可证为 MIT。128K 是部署规格，不是公布的上下文上限。模型卡没有发布日期，也没有第一方托管价格。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Ming-3.0-audio-flash model card | inclusionAI",
+      "url": "https://huggingface.co/inclusionAI/Ming-3.0-audio-flash",
+      "retrieved_at": "2026-10-24"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/models/ming-3.0-audio-tiny.json b/data/models/ming-3.0-audio-tiny.json
new file mode 100644
index 00000000..e3a3eb63
--- /dev/null
+++ b/data/models/ming-3.0-audio-tiny.json
@@ -0,0 +1,57 @@
+{
+  "id": "ming-3.0-audio-tiny",
+  "object": "model",
+  "created": null,
+  "owned_by": "inclusionai",
+  "name": "Ming-3.0-audio-tiny",
+  "api": {
+    "types": [
+      "chat",
+      "audio"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "audio"
+    ],
+    "output": [
+      "text",
+      "audio"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": []
+  },
+  "pricing": null,
+  "features": [],
+  "info": {
+    "status": "active",
+    "release_date": null,
+    "knowledge_cutoff": null,
+    "description": "inclusionAI compact native audio-language mixture-of-experts model with 9.3B total parameters and 2.7B activated per step. It takes audio and text and can return text or text with synthesized speech. The card states audio reasoning and the MIT license. No context limit, release date, or first-party hosted price is published.",
+    "docs": "https://huggingface.co/inclusionAI/Ming-3.0-audio-tiny",
+    "verified_at": "2026-10-24"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "inclusionAI 的紧凑原生音频语言模型，混合专家架构，总参数 9.3B，每步激活 2.7B。输入为音频和文本，可以返回文本，或文本加合成语音。模型卡写明支持音频推理，许可证为 MIT。未公布上下文上限、发布日期或第一方托管价格。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Ming-3.0-audio-tiny model card | inclusionAI",
+      "url": "https://huggingface.co/inclusionAI/Ming-3.0-audio-tiny",
+      "retrieved_at": "2026-10-24"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/models/ming-3.0-tts-0.5b.json b/data/models/ming-3.0-tts-0.5b.json
new file mode 100644
index 00000000..c38b186b
--- /dev/null
+++ b/data/models/ming-3.0-tts-0.5b.json
@@ -0,0 +1,57 @@
+{
+  "id": "ming-3.0-tts-0.5b",
+  "object": "model",
+  "created": null,
+  "owned_by": "inclusionai",
+  "name": "Ming-3.0-tts-0.5B",
+  "api": {
+    "types": [
+      "audio"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "audio"
+    ],
+    "output": [
+      "audio"
+    ]
+  },
+  "reasoning": {
+    "supported": null,
+    "efforts": []
+  },
+  "pricing": null,
+  "features": [],
+  "info": {
+    "status": "active",
+    "release_date": null,
+    "knowledge_cutoff": null,
+    "description": "inclusionAI dense 0.5B speech-synthesis checkpoint built on Qwen2.5-0.5B. It takes text and an optional reference clip and writes 44.1 kHz audio. The card lists voice cloning, voice design, emotion and dialect control, and the MIT license. No context limit, release date, or first-party hosted price is published.",
+    "docs": "https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B",
+    "verified_at": "2026-10-24"
+  },
+  "aliases": [
+    "Ming-3.0-tts-0.5B"
+  ],
+  "translations": {
+    "zh": {
+      "description": "inclusionAI 的稠密 0.5B 语音合成模型，骨干为 Qwen2.5-0.5B。输入为文本和可选的参考音频，输出 44.1 kHz 音频。模型卡列出声音克隆、声音设计、情绪和方言控制，许可证为 MIT。未公布上下文上限、发布日期或第一方托管价格。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Ming-3.0-tts-0.5B model card | inclusionAI",
+      "url": "https://huggingface.co/inclusionAI/Ming-3.0-tts-0.5B",
+      "retrieved_at": "2026-10-24"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/models/ming-3.0-tts-tiny.json b/data/models/ming-3.0-tts-tiny.json
new file mode 100644
index 00000000..83935ca6
--- /dev/null
+++ b/data/models/ming-3.0-tts-tiny.json
@@ -0,0 +1,57 @@
+{
+  "id": "ming-3.0-tts-tiny",
+  "object": "model",
+  "created": null,
+  "owned_by": "inclusionai",
+  "name": "Ming-3.0-tts-tiny",
+  "api": {
+    "types": [
+      "audio"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "audio"
+    ],
+    "output": [
+      "audio"
+    ]
+  },
+  "reasoning": {
+    "supported": null,
+    "efforts": []
+  },
+  "pricing": null,
+  "features": [],
+  "info": {
+    "status": "active",
+    "release_date": null,
+    "knowledge_cutoff": null,
+    "description": "inclusionAI mixture-of-experts speech-synthesis model. The Ling V3 backbone has 7.9B total parameters and about 1.3B active per token. It takes text and an optional reference clip and writes audio, including multi-speaker dialogue. The card lists the MIT license. No context limit, release date, or first-party hosted price is published.",
+    "docs": "https://huggingface.co/inclusionAI/Ming-3.0-tts-tiny",
+    "verified_at": "2026-10-24"
+  },
+  "aliases": [
+    "Ming-3.0-tts-tiny"
+  ],
+  "translations": {
+    "zh": {
+      "description": "inclusionAI 的混合专家语音合成模型。Ling V3 骨干总参数 7.9B，每 token 大约激活 1.3B。输入为文本和可选的参考音频，输出音频，并支持多说话人对话。模型卡写明许可证为 MIT。未公布上下文上限、发布日期或第一方托管价格。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Ming-3.0-tts-tiny model card | inclusionAI",
+      "url": "https://huggingface.co/inclusionAI/Ming-3.0-tts-tiny",
+      "retrieved_at": "2026-10-24"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/models/minicpm-v-4.7-1b.json b/data/models/minicpm-v-4.7-1b.json
new file mode 100644
index 00000000..fd4bd5df
--- /dev/null
+++ b/data/models/minicpm-v-4.7-1b.json
@@ -0,0 +1,58 @@
+{
+  "id": "minicpm-v-4.7-1b",
+  "object": "model",
+  "created": null,
+  "owned_by": "openbmb",
+  "name": "MiniCPM-V 4.7 1B",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "image",
+      "video"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": []
+  },
+  "pricing": null,
+  "features": [],
+  "info": {
+    "status": "active",
+    "release_date": null,
+    "knowledge_cutoff": null,
+    "description": "Open-weight multimodal model with 1.55B parameters. The docs list a 256K context and support for single-image, multi-image, and video understanding. Thinking is switched with enable_thinking; no effort values are listed. OpenBMB is not in the abbreviation table, so 256K is not converted. The page does not publish a calendar date, a maximum output, a standard API price, or a first-party request URL.",
+    "docs": "https://openbmb.cn/docs/en/guides/minicpm-v/minicpm-v-4.7",
+    "verified_at": "2026-10-27"
+  },
+  "aliases": [
+    "openbmb/MiniCPM-V-4.7-1B"
+  ],
+  "translations": {
+    "zh": {
+      "description": "开放权重多模态模型，参数 1.55B。文档列出 256K 上下文，并支持单图、多图和视频理解。思考由 enable_thinking 开关控制，没有列出强度取值。OpenBMB 不在缩写换算表中，因此不换算 256K。页面没有给出日历日期、最大输出、API 标准价或第一方请求 URL。"
+    }
+  },
+  "sources": [
+    {
+      "title": "MiniCPM-V 4.7 | OpenBMB",
+      "url": "https://openbmb.cn/docs/en/guides/minicpm-v/minicpm-v-4.7",
+      "retrieved_at": "2026-10-27"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/models/minicpm-v-4.7-35b-a3b.json b/data/models/minicpm-v-4.7-35b-a3b.json
new file mode 100644
index 00000000..2676c59f
--- /dev/null
+++ b/data/models/minicpm-v-4.7-35b-a3b.json
@@ -0,0 +1,58 @@
+{
+  "id": "minicpm-v-4.7-35b-a3b",
+  "object": "model",
+  "created": null,
+  "owned_by": "openbmb",
+  "name": "MiniCPM-V 4.7 35B-A3B",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "image",
+      "video"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": []
+  },
+  "pricing": null,
+  "features": [],
+  "info": {
+    "status": "active",
+    "release_date": null,
+    "knowledge_cutoff": null,
+    "description": "Open-weight multimodal model with 35.2B total parameters and about 3.5B activated. The docs list a 256K context and support for single-image, multi-image, and video understanding. Thinking is switched with enable_thinking; no effort values are listed. OpenBMB is not in the abbreviation table, so 256K is not converted. The page does not publish a calendar date, a maximum output, a standard API price, or a first-party request URL.",
+    "docs": "https://openbmb.cn/docs/en/guides/minicpm-v/minicpm-v-4.7",
+    "verified_at": "2026-10-27"
+  },
+  "aliases": [
+    "openbmb/MiniCPM-V-4.7-35B-A3B"
+  ],
+  "translations": {
+    "zh": {
+      "description": "开放权重多模态模型，总参数 35.2B，激活约 3.5B。文档列出 256K 上下文，并支持单图、多图和视频理解。思考由 enable_thinking 开关控制，没有列出强度取值。OpenBMB 不在缩写换算表中，因此不换算 256K。页面没有给出日历日期、最大输出、API 标准价或第一方请求 URL。"
+    }
+  },
+  "sources": [
+    {
+      "title": "MiniCPM-V 4.7 | OpenBMB",
+      "url": "https://openbmb.cn/docs/en/guides/minicpm-v/minicpm-v-4.7",
+      "retrieved_at": "2026-10-27"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/models/minimax-m3.1-flash.json b/data/models/minimax-m3.1-flash.json
new file mode 100644
index 00000000..de1c24e8
--- /dev/null
+++ b/data/models/minimax-m3.1-flash.json
@@ -0,0 +1,114 @@
+{
+  "id": "minimax-m3.1-flash",
+  "object": "model",
+  "created": 1792368000,
+  "owned_by": "minimax",
+  "name": "MiniMax-M3.1-Flash",
+  "api": {
+    "types": [
+      "chat",
+      "messages"
+    ],
+    "sdk": [
+      "@anthropic-ai/sdk",
+      "openai"
+    ]
+  },
+  "limits": {
+    "context": 1000000,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "image",
+      "video"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": [
+      "low",
+      "medium",
+      "high",
+      "xhigh",
+      "max"
+    ]
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "1M_tokens",
+    "input": 0.15,
+    "output": 0.6,
+    "cache_read": 0.015,
+    "cache_write": null
+  },
+  "features": [
+    "tools",
+    "streaming",
+    "caching"
+  ],
+  "info": {
+    "status": "active",
+    "release_date": "2026-10-19",
+    "knowledge_cutoff": null,
+    "description": "Multimodal agent model for coding and tool use. Developer docs list a 1,000,000-token context window, text, image, and video input, and text output. Effort accepts low, medium, high, xhigh, and max, defaults to max, and cannot be disabled. Hugging Face publishes the weights at approximately 428B total parameters with 23B activated per token, under the MiniMax Community License: non-commercial use is free; commercial use must display “Built with MiniMax M3.1-Flash”; yearly revenue above $20 million USD requires prior written authorization from MiniMax, and otherwise a one-time notice.",
+    "docs": "https://platform.minimax.io/docs/guides/text-generation",
+    "verified_at": "2026-10-24"
+  },
+  "aliases": [
+    "MiniMax-M3.1-Flash"
+  ],
+  "translations": {
+    "zh": {
+      "description": "面向编程和工具调用的多模态智能体模型。开发者文档写明上下文窗口为 1,000,000 tokens，输入支持文本、图像和视频，输出为文本。effort 取值为 low、medium、high、xhigh 和 max，默认 max，且不能关闭思考。Hugging Face 公布的权重大约总参数 428B、每 token 激活 23B，许可证为 MiniMax Community License：非商业使用免费；商业使用须标明 “Built with MiniMax M3.1-Flash”；年收入超过 2000 万美元须事先取得 MiniMax 书面授权，否则发送一次通知即可。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Model Invocation | MiniMax API Docs",
+      "url": "https://platform.minimax.io/docs/guides/text-generation",
+      "retrieved_at": "2026-10-22"
+    },
+    {
+      "title": "Pay as You Go pricing | MiniMax API Docs",
+      "url": "https://platform.minimax.io/docs/guides/pricing-paygo",
+      "retrieved_at": "2026-10-22"
+    },
+    {
+      "title": "MiniMax M3.1 Flash | MiniMax",
+      "url": "https://www.minimax.io/blog/minimax-m31f",
+      "retrieved_at": "2026-10-22"
+    },
+    {
+      "title": "MiniMax-M3.1-Flash model card | MiniMax",
+      "url": "https://huggingface.co/MiniMaxAI/MiniMax-M3.1-Flash",
+      "retrieved_at": "2026-10-24"
+    },
+    {
+      "title": "MiniMax Community License | MiniMax-M3.1-Flash",
+      "url": "https://huggingface.co/MiniMaxAI/MiniMax-M3.1-Flash/blob/main/LICENSE",
+      "retrieved_at": "2026-10-24"
+    }
+  ],
+  "providers": [
+    {
+      "id": "minimax",
+      "model": "MiniMax-M3.1-Flash",
+      "base_url": "https://api.minimax.io",
+      "endpoints": {
+        "chat": "POST /v1/chat/completions",
+        "messages": "POST /anthropic/v1/messages"
+      },
+      "reasoning_param": "reasoning_effort",
+      "reasoning_params": {
+        "chat": "reasoning_effort",
+        "messages": "output_config.effort"
+      }
+    }
+  ]
+}
diff --git a/data/models/moondream3-preview.json b/data/models/moondream3-preview.json
new file mode 100644
index 00000000..c1ba50ff
--- /dev/null
+++ b/data/models/moondream3-preview.json
@@ -0,0 +1,80 @@
+{
+  "id": "moondream3-preview",
+  "object": "model",
+  "created": null,
+  "owned_by": "moondream",
+  "name": "Moondream 3 Preview",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "image"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": null,
+    "efforts": []
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "1M_tokens",
+    "input": 0.3,
+    "output": 2.5,
+    "cache_read": null,
+    "cache_write": null
+  },
+  "features": [
+    "streaming",
+    "batch"
+  ],
+  "info": {
+    "status": "preview",
+    "release_date": null,
+    "knowledge_cutoff": null,
+    "description": "Hosted Moondream 3 Preview vision model. The API model id is moondream3-preview, and the shared chat endpoint accepts image inputs and streaming. Real-time price is $0.30 per million input tokens and $2.50 per million output tokens. The batch price is half of that and is not stored. Context length, parameter count, and release date are not stated on the pricing or API pages.",
+    "docs": "https://docs.moondream.ai/api",
+    "verified_at": "2026-10-24"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "托管的 Moondream 3 Preview 视觉模型。API 模型 ID 为 moondream3-preview，共用的对话端点接受图像输入并支持流式输出。实时价为每百万输入 token 0.30 美元、每百万输出 token 2.50 美元。批量价是该价格的一半，未记入。价格页和 API 页没有写上下文长度、参数量和发布日期。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Pricing | Moondream",
+      "url": "https://moondream.ai/pricing",
+      "retrieved_at": "2026-10-24"
+    },
+    {
+      "title": "API | Moondream Docs",
+      "url": "https://docs.moondream.ai/api",
+      "retrieved_at": "2026-10-24"
+    }
+  ],
+  "providers": [
+    {
+      "id": "moondream",
+      "model": "moondream3-preview",
+      "base_url": "https://api.moondream.ai",
+      "endpoints": {
+        "chat": "POST /v1/chat/completions"
+      },
+      "reasoning_param": null
+    }
+  ]
+}
diff --git a/data/models/moondream3.1-9b-a2b.json b/data/models/moondream3.1-9b-a2b.json
new file mode 100644
index 00000000..21452702
--- /dev/null
+++ b/data/models/moondream3.1-9b-a2b.json
@@ -0,0 +1,92 @@
+{
+  "id": "moondream3.1-9b-a2b",
+  "object": "model",
+  "created": 1783382400,
+  "owned_by": "moondream",
+  "name": "Moondream 3.1 9B A2B",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": 32768,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "image"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": []
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "1M_tokens",
+    "input": 0.3,
+    "output": 1,
+    "cache_read": null,
+    "cache_write": null
+  },
+  "features": [
+    "streaming",
+    "batch"
+  ],
+  "info": {
+    "status": "active",
+    "release_date": "2026-07-07",
+    "knowledge_cutoff": "2025-12",
+    "description": "Open-weight vision-language model with 9.0B total parameters and 2.0B active per token. It reads image and text into a 32,768-token context, returns text, and supports grounded reasoning. Weights use the Moondream License. Real-time cloud price is $0.30 per million input tokens and $1.00 per million output tokens. The batch price is half of that and is not stored. The model card still labels its figures preliminary.",
+    "docs": "https://moondream.ai/models/moondream_3-1_9B_A2B",
+    "verified_at": "2026-10-24"
+  },
+  "aliases": [
+    "moondream3.1-9B-A2B"
+  ],
+  "translations": {
+    "zh": {
+      "description": "开放权重的视觉语言模型，总参数 9.0B，每 token 激活 2.0B。它把图像和文本读入 32,768 token 上下文，输出文本，并支持带定位的推理。权重使用 Moondream License。云端实时价为每百万输入 token 0.30 美元、每百万输出 token 1.00 美元。批量价是该价格的一半，未记入。模型卡仍把数字标为初步结果。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Moondream 3.1 9B A2B model card",
+      "url": "https://moondream.ai/models/moondream_3-1_9B_A2B",
+      "retrieved_at": "2026-10-24"
+    },
+    {
+      "title": "Pricing | Moondream",
+      "url": "https://moondream.ai/pricing",
+      "retrieved_at": "2026-10-24"
+    },
+    {
+      "title": "API | Moondream Docs",
+      "url": "https://docs.moondream.ai/api",
+      "retrieved_at": "2026-10-24"
+    },
+    {
+      "title": "Moondream 3.1: Beyond Benchmarks",
+      "url": "https://moondream.ai/blog/moondream-3-1-beyond-benchmarks",
+      "retrieved_at": "2026-10-24"
+    }
+  ],
+  "providers": [
+    {
+      "id": "moondream",
+      "model": "moondream3.1-9B-A2B",
+      "base_url": "https://api.moondream.ai",
+      "endpoints": {
+        "chat": "POST /v1/chat/completions"
+      },
+      "reasoning_param": null
+    }
+  ]
+}
diff --git a/data/models/muse-spark-open.json b/data/models/muse-spark-open.json
new file mode 100644
index 00000000..4f48df0e
--- /dev/null
+++ b/data/models/muse-spark-open.json
@@ -0,0 +1,67 @@
+{
+  "id": "muse-spark-open",
+  "object": "model",
+  "created": 1792022400,
+  "owned_by": "meta",
+  "name": "Muse Spark Open",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "image",
+      "video",
+      "audio"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": [
+      "low",
+      "medium",
+      "high",
+      "xhigh"
+    ]
+  },
+  "pricing": null,
+  "features": [],
+  "info": {
+    "status": "active",
+    "release_date": "2026-10-15",
+    "knowledge_cutoff": null,
+    "description": "Open-weight Muse Spark model with 816B total parameters and 42B active, released under the Muse Model License. The model card states a 1M-token context at inference; Meta publishes exact token counts per model and this page does not, so the context limit is unset. Input is text, image, video, and audio, and output is text. reasoning_effort accepts low, medium, high, or xhigh, and the default is xhigh. Meta does not publish a first-party API or a standard API price; weights are on Hugging Face and inference is hosted by third parties.",
+    "docs": "https://research.meta.ai/blog/muse-spark-open",
+    "verified_at": "2026-10-16"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "开放权重的 Muse Spark 模型，总参数 816B、激活 42B，许可证为 Muse Model License。模型卡写推理时上下文为 1M tokens；Meta 按模型给出精确 token 数，该页没有精确数字，因此上下文上限留空。输入为文本、图像、视频和音频，输出为文本。reasoning_effort 可取 low、medium、high 或 xhigh，默认 xhigh。Meta 未公布第一方 API，也没有 API 标准价；权重在 Hugging Face，推理由第三方托管。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Introducing Muse Spark Open",
+      "url": "https://research.meta.ai/blog/muse-spark-open",
+      "retrieved_at": "2026-10-16"
+    },
+    {
+      "title": "Muse Spark Open model card（816B/42B、1M、模态与 reasoning_effort）| Hugging Face",
+      "url": "https://huggingface.co/meta-models/Muse-Spark-Open-816B-A42B",
+      "retrieved_at": "2026-10-16"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/models/muse-wonder-1.0.json b/data/models/muse-wonder-1.0.json
new file mode 100644
index 00000000..d871a859
--- /dev/null
+++ b/data/models/muse-wonder-1.0.json
@@ -0,0 +1,119 @@
+{
+  "id": "muse-wonder-1.0",
+  "object": "model",
+  "created": 1791763200,
+  "owned_by": "meta",
+  "name": "Muse Wonder 1.0",
+  "api": {
+    "types": [
+      "chat",
+      "responses"
+    ],
+    "sdk": [
+      "openai"
+    ]
+  },
+  "limits": {
+    "context": 1048576,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "image",
+      "video",
+      "audio"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": [
+      "minimal",
+      "low",
+      "medium",
+      "high",
+      "xhigh",
+      "max"
+    ]
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "1M_tokens",
+    "input": 4,
+    "output": 20,
+    "cache_read": 0.2,
+    "cache_write": null
+  },
+  "features": [
+    "tools",
+    "structured_output",
+    "streaming",
+    "caching"
+  ],
+  "info": {
+    "status": "preview",
+    "release_date": "2026-10-12",
+    "knowledge_cutoff": null,
+    "description": "Meta's most capable model, built for complex software engineering and long-horizon agentic work, with a 1,048,576-token context window.",
+    "docs": "https://dev.meta.ai/models/muse-wonder",
+    "verified_at": "2026-10-26"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "Meta 能力最强的模型，面向复杂软件工程和长周期智能体任务，上下文窗口为 1,048,576 tokens。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Muse Wonder | Meta Model API",
+      "url": "https://dev.meta.ai/models/muse-wonder",
+      "retrieved_at": "2026-10-26"
+    },
+    {
+      "title": "Build with Meta Model API and Muse Code | Meta Model API",
+      "url": "https://dev.meta.ai/docs/overview",
+      "retrieved_at": "2026-10-26"
+    },
+    {
+      "title": "Pricing and rate limits | Meta Model API",
+      "url": "https://dev.meta.ai/docs/pricing-rate-limits",
+      "retrieved_at": "2026-10-26"
+    },
+    {
+      "title": "Reasoning | Meta Model API",
+      "url": "https://dev.meta.ai/docs/reasoning",
+      "retrieved_at": "2026-10-26"
+    },
+    {
+      "title": "Muse Wonder Model Card | Meta",
+      "url": "https://research.meta.ai/static/muse-wonder-model-card",
+      "retrieved_at": "2026-10-26"
+    },
+    {
+      "title": "SDK support — openai (api.sdk only)",
+      "url": "https://github.com/openai/openai-node",
+      "retrieved_at": "2026-10-26"
+    }
+  ],
+  "providers": [
+    {
+      "id": "meta",
+      "model": "muse-wonder-1.0",
+      "base_url": "https://api.meta.ai/v1",
+      "endpoints": {
+        "chat": "POST /chat/completions",
+        "responses": "POST /responses"
+      },
+      "reasoning_param": "reasoning_effort",
+      "reasoning_params": {
+        "chat": "reasoning_effort",
+        "responses": "reasoning.effort"
+      }
+    }
+  ]
+}
diff --git a/data/models/nemotron-3.5-embed-1b.json b/data/models/nemotron-3.5-embed-1b.json
new file mode 100644
index 00000000..a9971448
--- /dev/null
+++ b/data/models/nemotron-3.5-embed-1b.json
@@ -0,0 +1,67 @@
+{
+  "id": "nemotron-3.5-embed-1b",
+  "object": "model",
+  "created": 1792454400,
+  "owned_by": "nvidia",
+  "name": "Nemotron 3.5 Embed 1B",
+  "api": {
+    "types": [
+      "embeddings"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": 32768,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "image"
+    ],
+    "output": [
+      "embedding"
+    ]
+  },
+  "reasoning": {
+    "supported": null,
+    "efforts": []
+  },
+  "pricing": null,
+  "features": [],
+  "info": {
+    "status": "preview",
+    "release_date": "2026-10-20",
+    "knowledge_cutoff": null,
+    "description": "1.55B multimodal embedding model. It embeds text and document images into a 2048-dimensional vector, with a 32,768-token context. The hosted Preview API serves the BF16 variant as nvidia/nemotron-3.5-embed-1b. Weights use OpenMDW-1.1. The same card also describes an NVFP4 quantized derivative that is not a separate hosted model id. The trial endpoint does not publish a numeric standard price.",
+    "docs": "https://build.nvidia.com/nvidia/nemotron-3_5-embed-1b/modelcard",
+    "verified_at": "2026-10-24"
+  },
+  "aliases": [
+    "nvidia/nemotron-3.5-embed-1b"
+  ],
+  "translations": {
+    "zh": {
+      "description": "1.55B 多模态向量模型。它把文本和文档图像编码为 2048 维向量，上下文 32,768 tokens。托管的 Preview API 提供 BF16 变体，模型 ID 为 nvidia/nemotron-3.5-embed-1b。权重使用 OpenMDW-1.1。同一模型卡还描述了 NVFP4 量化衍生版本，它不是另一个托管模型 ID。试用端点没有公布数字标准价。"
+    }
+  },
+  "sources": [
+    {
+      "title": "nemotron-3.5-embed-1b Model Card | NVIDIA NIM",
+      "url": "https://build.nvidia.com/nvidia/nemotron-3_5-embed-1b/modelcard",
+      "retrieved_at": "2026-10-24"
+    }
+  ],
+  "providers": [
+    {
+      "id": "nvidia",
+      "model": "nvidia/nemotron-3.5-embed-1b",
+      "base_url": "https://integrate.api.nvidia.com",
+      "endpoints": {
+        "embeddings": "POST /v1/embeddings"
+      },
+      "reasoning_param": null
+    }
+  ]
+}
diff --git a/data/models/nemotron-3.5-embed-8b.json b/data/models/nemotron-3.5-embed-8b.json
new file mode 100644
index 00000000..94035a73
--- /dev/null
+++ b/data/models/nemotron-3.5-embed-8b.json
@@ -0,0 +1,62 @@
+{
+  "id": "nemotron-3.5-embed-8b",
+  "object": "model",
+  "created": null,
+  "owned_by": "nvidia",
+  "name": "Nemotron 3.5 Embed 8B",
+  "api": {
+    "types": [
+      "embeddings"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": 32768,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "image"
+    ],
+    "output": [
+      "embedding"
+    ]
+  },
+  "reasoning": {
+    "supported": null,
+    "efforts": []
+  },
+  "pricing": null,
+  "features": [],
+  "info": {
+    "status": "preview",
+    "release_date": null,
+    "knowledge_cutoff": null,
+    "description": "8.38B multimodal embedding model for text and document images, with a 32,768-token context and 4096-dimensional embeddings. Weights use OpenMDW-1.1. The same card lists a Build.NVIDIA.com release of 2026-10-19 and a Hugging Face release of 2026-10-20, so the release date is unset. No request path is printed, and the trial endpoint does not publish a numeric standard price.",
+    "docs": "https://build.nvidia.com/nvidia/nemotron-3_5-embed-8b/modelcard",
+    "verified_at": "2026-10-24"
+  },
+  "aliases": [
+    "nvidia/nemotron-3.5-embed-8b"
+  ],
+  "translations": {
+    "zh": {
+      "description": "8.38B 多模态向量模型，输入为文本和文档图像，上下文 32,768 tokens，向量维度 4096。权重使用 OpenMDW-1.1。同一模型卡同时写了 Build.NVIDIA.com 发布日 2026-10-19 和 Hugging Face 发布日 2026-10-20，因此发布日期留空。页面没有给出请求路径，试用端点也没有数字标准价。"
+    }
+  },
+  "sources": [
+    {
+      "title": "nemotron-3.5-embed-8b Model Card | NVIDIA NIM",
+      "url": "https://build.nvidia.com/nvidia/nemotron-3_5-embed-8b/modelcard",
+      "retrieved_at": "2026-10-24"
+    },
+    {
+      "title": "nvidia / nemotron-3.5-embed-8b | NVIDIA NIM API Reference",
+      "url": "https://docs.api.nvidia.com/nim/reference/nvidia-nemotron-3_5-embed-8b",
+      "retrieved_at": "2026-10-24"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/models/nemotron-3.5-rerank-8b.json b/data/models/nemotron-3.5-rerank-8b.json
new file mode 100644
index 00000000..26abb210
--- /dev/null
+++ b/data/models/nemotron-3.5-rerank-8b.json
@@ -0,0 +1,57 @@
+{
+  "id": "nemotron-3.5-rerank-8b",
+  "object": "model",
+  "created": 1792454400,
+  "owned_by": "nvidia",
+  "name": "Nemotron 3.5 Rerank 8B",
+  "api": {
+    "types": [
+      "rerank"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": 32768,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "image"
+    ],
+    "output": [
+      "score"
+    ]
+  },
+  "reasoning": {
+    "supported": null,
+    "efforts": []
+  },
+  "pricing": null,
+  "features": [],
+  "info": {
+    "status": "preview",
+    "release_date": "2026-10-20",
+    "knowledge_cutoff": null,
+    "description": "8.38B reranker that scores a text query against a document of text, a page image, or both. Context is 32,768 tokens. Weights use OpenMDW-1.1. The API reference dates both the Hugging Face card and the planned Build.NVIDIA listing to 2026-10-20. No request path is printed, and no numeric standard price is published.",
+    "docs": "https://docs.api.nvidia.com/nim/reference/nvidia-nemotron-3_5-rerank-8b",
+    "verified_at": "2026-10-24"
+  },
+  "aliases": [
+    "nvidia/nemotron-3.5-rerank-8b"
+  ],
+  "translations": {
+    "zh": {
+      "description": "8.38B 重排模型，对文本查询与文档（文本、页面图像或两者）的相关程度打分。上下文 32,768 tokens。权重使用 OpenMDW-1.1。API 参考把 Hugging Face 模型卡和计划中的 Build.NVIDIA 上架都写为 2026-10-20。页面没有给出请求路径，也没有数字标准价。"
+    }
+  },
+  "sources": [
+    {
+      "title": "nvidia / nemotron-3.5-rerank-8b | NVIDIA NIM API Reference",
+      "url": "https://docs.api.nvidia.com/nim/reference/nvidia-nemotron-3_5-rerank-8b",
+      "retrieved_at": "2026-10-24"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/models/nemotron-3.5-super-vl.json b/data/models/nemotron-3.5-super-vl.json
new file mode 100644
index 00000000..89453675
--- /dev/null
+++ b/data/models/nemotron-3.5-super-vl.json
@@ -0,0 +1,88 @@
+{
+  "id": "nemotron-3.5-super-vl",
+  "object": "model",
+  "created": 1792022400,
+  "owned_by": "nvidia",
+  "name": "Nemotron 3.5 Super VL",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": [
+      "openai"
+    ]
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "image",
+      "video"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": []
+  },
+  "pricing": null,
+  "features": [
+    "tools",
+    "streaming"
+  ],
+  "info": {
+    "status": "active",
+    "release_date": "2026-10-15",
+    "knowledge_cutoff": null,
+    "description": "Open 120B-parameter vision-language model with 12B active parameters, under the OpenMDW-1.1 license. It accepts text, images, and video and returns text; the announcement also describes webpage and PDF understanding, with documents supplied as page images. The weights card says context is up to 1M tokens, and one TensorRT-LLM example sets max_seq_len to 1048576. The NVIDIA NIM catalog lists 262K for the hosted endpoint. Those figures conflict, so the context limit is unset. Reasoning is switched with chat_template_kwargs.enable_thinking; medium effort sets medium_effort to true. No official API token price is published.",
+    "docs": "https://blogs.nvidia.com/blog/nemotron-3-5-super-vl/",
+    "verified_at": "2026-10-16"
+  },
+  "aliases": [
+    "nvidia/nemotron-3.5-super-vl-120b-a12b"
+  ],
+  "translations": {
+    "zh": {
+      "description": "开放的 120B 视觉语言模型，激活参数 12B，许可证为 OpenMDW-1.1。输入为文本、图像和视频，输出为文本；公告还写到网页与 PDF 理解，文档以页面图像提供。权重卡写上下文最高 1M tokens，一份 TensorRT-LLM 示例把 max_seq_len 设为 1048576。NVIDIA NIM 目录把托管端点标为 262K。两处数字冲突，因此上下文上限留空。推理用 chat_template_kwargs.enable_thinking 开关；中等强度另设 medium_effort 为 true。官方未公布 API token 标准价。"
+    }
+  },
+  "sources": [
+    {
+      "title": "NVIDIA Expands Nemotron Family With Nemotron 3.5 Super VL",
+      "url": "https://blogs.nvidia.com/blog/nemotron-3-5-super-vl/",
+      "retrieved_at": "2026-10-16"
+    },
+    {
+      "title": "NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4 model card | Hugging Face",
+      "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4",
+      "retrieved_at": "2026-10-16"
+    },
+    {
+      "title": "nemotron-3.5-super-vl-120b-a12b model card | NVIDIA NIM",
+      "url": "https://build.nvidia.com/nvidia/nemotron-3.5-super-vl-120b-a12b/modelcard",
+      "retrieved_at": "2026-10-16"
+    },
+    {
+      "title": "nemotron-3.5-super-vl-120b-a12b | NVIDIA NIM",
+      "url": "https://build.nvidia.com/nvidia/nemotron-3.5-super-vl-120b-a12b",
+      "retrieved_at": "2026-10-16"
+    }
+  ],
+  "providers": [
+    {
+      "id": "nvidia",
+      "model": "nvidia/nemotron-3.5-super-vl-120b-a12b",
+      "base_url": "https://integrate.api.nvidia.com",
+      "endpoints": {
+        "chat": "POST /v1/chat/completions"
+      },
+      "reasoning_param": null
+    }
+  ]
+}
diff --git a/data/models/north-small-translate-1-0.json b/data/models/north-small-translate-1-0.json
index 5ae0cb32..ecc1fcf4 100644
--- a/data/models/north-small-translate-1-0.json
+++ b/data/models/north-small-translate-1-0.json
@@ -38,9 +38,11 @@
     "knowledge_cutoff": null,
     "description": "218B total / 25B active parameter MoE model for machine translation across 50+ languages.",
     "docs": "https://docs.cohere.com/docs/north-small-translate-1.0",
-    "verified_at": "2026-10-11"
+    "verified_at": "2026-10-26"
   },
-  "aliases": [],
+  "aliases": [
+    "north-small-translate-09-2026"
+  ],
   "translations": {
     "zh": {
       "description": "总参数 218B、激活参数 25B 的 MoE 模型，用于 50 多种语言的机器翻译。"
@@ -50,12 +52,12 @@
     {
       "title": "North Small Translate 1.0 | Cohere Docs",
       "url": "https://docs.cohere.com/docs/north-small-translate-1.0",
-      "retrieved_at": "2026-10-10"
+      "retrieved_at": "2026-10-26"
     },
     {
       "title": "Cohere Model Overview",
       "url": "https://docs.cohere.com/docs/models",
-      "retrieved_at": "2026-10-10"
+      "retrieved_at": "2026-10-26"
     },
     {
       "title": "How does Cohere pricing work?",
@@ -81,7 +83,7 @@
   "providers": [
     {
       "id": "cohere",
-      "model": "north-small-translate-1-0",
+      "model": "north-small-translate-09-2026",
       "base_url": "https://api.cohere.com",
       "endpoints": {
         "chat": "POST /v2/chat"
diff --git a/data/models/pareto-26.10.json b/data/models/pareto-26.10.json
new file mode 100644
index 00000000..306af06b
--- /dev/null
+++ b/data/models/pareto-26.10.json
@@ -0,0 +1,65 @@
+{
+  "id": "pareto-26.10",
+  "object": "model",
+  "created": 1792022400,
+  "owned_by": "unbiased",
+  "name": "Pareto 26.10",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": null,
+    "efforts": []
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "1M_tokens",
+    "input": 0.8,
+    "output": 3.2,
+    "cache_read": 0.03,
+    "cache_write": null
+  },
+  "features": [
+    "caching"
+  ],
+  "info": {
+    "status": "active",
+    "release_date": "2026-10-15",
+    "knowledge_cutoff": null,
+    "description": "Stable Pareto release as of October 15, 2026. On the Unbiased API, pareto points at 26.10, and the dated id pareto-26.10 pins this release. The announcement says the preview price is now permanent: $0.80 per million input tokens, $0.03 cached, and $3.20 output. It also says the context window is 1M tokens. Unbiased is not in the abbreviation table and the page does not print an exact token count, so the context limit is unset. No cache-write price, maximum output, or request URL is published. Subscriber weekly allowances are plan prices and are not recorded.",
+    "docs": "https://unbiased.ai/blog/pareto-26-10/",
+    "verified_at": "2026-10-27"
+  },
+  "aliases": [
+    "pareto"
+  ],
+  "translations": {
+    "zh": {
+      "description": "2026 年 10 月 15 日起的稳定版 Pareto。在 Unbiased API 上，pareto 指向 26.10，带日期的 id pareto-26.10 用来固定这个版本。公告写明预览价现已转为永久价：输入每百万 tokens 0.80 美元，缓存 0.03 美元，输出 3.20 美元。公告还写上下文窗口为 1M tokens。Unbiased 不在缩写换算表中，页面也没有精确 token 数，因此上下文上限留空。没有公布缓存写入价、最大输出或请求 URL。订阅套餐的周额度是方案价格，不记录。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Pareto 26.10 is out of preview | Unbiased AI",
+      "url": "https://unbiased.ai/blog/pareto-26-10/",
+      "retrieved_at": "2026-10-27"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/models/perceptron-mk1.5.json b/data/models/perceptron-mk1.5.json
new file mode 100644
index 00000000..ddd6ed4a
--- /dev/null
+++ b/data/models/perceptron-mk1.5.json
@@ -0,0 +1,66 @@
+{
+  "id": "perceptron-mk1.5",
+  "object": "model",
+  "created": null,
+  "owned_by": "perceptron",
+  "name": "Perceptron Mk1.5",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "image",
+      "audio",
+      "video"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": null,
+    "efforts": []
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "1M_tokens",
+    "input": 0.15,
+    "output": 1.5,
+    "cache_read": null,
+    "cache_write": null
+  },
+  "features": [
+    "tools"
+  ],
+  "info": {
+    "status": "active",
+    "release_date": null,
+    "knowledge_cutoff": null,
+    "description": "Mk1.5 ingests text, images, video, and audio and emits text, points, boxes, polygons, clips, and object tracks. The API model ID is perceptron-mk1.5. The page lists a 32K-token multimodal context window and prices of $0.15 per 1M input tokens and $1.50 per 1M output tokens. It supports arbitrary tools in OpenAI JSON Schema format. Perceptron is not in the abbreviation table, so 32K is not converted. The page does not publish a calendar date, a cached-token price, a maximum output, or a request URL.",
+    "docs": "https://www.perceptron.inc/blog/introducing-perceptron-mk1-5",
+    "verified_at": "2026-10-27"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "Mk1.5 接收文本、图像、视频和音频，输出文本以及点、框、多边形、片段和对象轨迹。API 模型 ID 为 perceptron-mk1.5。页面写明 32K tokens 的多模态上下文，价格为输入每百万 tokens 0.15 美元、输出 1.50 美元。它支持 OpenAI JSON Schema 格式的任意工具。Perceptron 不在缩写换算表中，因此不把 32K 换算成整数。页面没有给出日历日期、缓存价、最大输出或请求 URL。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Introducing Perceptron Mk1.5",
+      "url": "https://www.perceptron.inc/blog/introducing-perceptron-mk1-5",
+      "retrieved_at": "2026-10-27"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/models/rerank-v5.0-fast.json b/data/models/rerank-v5.0-fast.json
new file mode 100644
index 00000000..1bb41e8f
--- /dev/null
+++ b/data/models/rerank-v5.0-fast.json
@@ -0,0 +1,76 @@
+{
+  "id": "rerank-v5.0-fast",
+  "object": "model",
+  "created": 1792540800,
+  "owned_by": "cohere",
+  "name": "Rerank v5.0 Fast",
+  "api": {
+    "types": [
+      "rerank"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": 32768,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text"
+    ],
+    "output": [
+      "score"
+    ]
+  },
+  "reasoning": {
+    "supported": null,
+    "efforts": []
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "1M_tokens",
+    "input": 0.02,
+    "output": null,
+    "cache_read": null,
+    "cache_write": null
+  },
+  "features": [],
+  "info": {
+    "status": "active",
+    "release_date": "2026-10-21",
+    "knowledge_cutoff": null,
+    "description": "Rerank 5 Fast ranks documents against a query. The docs state a context length of 32,768 tokens for the query and documents combined. The launch post prices the API at $0.02 per 1M tokens. No cached-token price, separate input cap, or output cap is published.",
+    "docs": "https://docs.cohere.com/docs/rerank",
+    "verified_at": "2026-10-27"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "Rerank 5 Fast 按查询对文档排序。文档写明查询与文档合计的上下文长度为 32,768 tokens。发布说明中的 API 价格为每百万 tokens 0.02 美元。页面没有给出缓存价、单独的输入上限或输出上限。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Rerank 5 | Cohere",
+      "url": "https://cohere.com/blog/rerank-5",
+      "retrieved_at": "2026-10-27"
+    },
+    {
+      "title": "Rerank | Cohere Docs",
+      "url": "https://docs.cohere.com/docs/rerank",
+      "retrieved_at": "2026-10-27"
+    }
+  ],
+  "providers": [
+    {
+      "id": "cohere",
+      "model": "rerank-v5.0-fast",
+      "base_url": "https://api.cohere.com",
+      "endpoints": {
+        "rerank": "POST /v2/rerank"
+      },
+      "reasoning_param": null
+    }
+  ]
+}
diff --git a/data/models/rerank-v5.0-pro.json b/data/models/rerank-v5.0-pro.json
new file mode 100644
index 00000000..94dd66de
--- /dev/null
+++ b/data/models/rerank-v5.0-pro.json
@@ -0,0 +1,76 @@
+{
+  "id": "rerank-v5.0-pro",
+  "object": "model",
+  "created": 1792540800,
+  "owned_by": "cohere",
+  "name": "Rerank v5.0 Pro",
+  "api": {
+    "types": [
+      "rerank"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": 32768,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text"
+    ],
+    "output": [
+      "score"
+    ]
+  },
+  "reasoning": {
+    "supported": null,
+    "efforts": []
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "1M_tokens",
+    "input": 0.05,
+    "output": null,
+    "cache_read": null,
+    "cache_write": null
+  },
+  "features": [],
+  "info": {
+    "status": "active",
+    "release_date": "2026-10-21",
+    "knowledge_cutoff": null,
+    "description": "Rerank 5 Pro ranks documents against a query. The docs state a context length of 32,768 tokens for the query and documents combined. The launch post prices the API at $0.05 per 1M tokens. No cached-token price, separate input cap, or output cap is published.",
+    "docs": "https://docs.cohere.com/docs/rerank",
+    "verified_at": "2026-10-27"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "Rerank 5 Pro 按查询对文档排序。文档写明查询与文档合计的上下文长度为 32,768 tokens。发布说明中的 API 价格为每百万 tokens 0.05 美元。页面没有给出缓存价、单独的输入上限或输出上限。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Rerank 5 | Cohere",
+      "url": "https://cohere.com/blog/rerank-5",
+      "retrieved_at": "2026-10-27"
+    },
+    {
+      "title": "Rerank | Cohere Docs",
+      "url": "https://docs.cohere.com/docs/rerank",
+      "retrieved_at": "2026-10-27"
+    }
+  ],
+  "providers": [
+    {
+      "id": "cohere",
+      "model": "rerank-v5.0-pro",
+      "base_url": "https://api.cohere.com",
+      "endpoints": {
+        "rerank": "POST /v2/rerank"
+      },
+      "reasoning_param": null
+    }
+  ]
+}
diff --git a/data/models/spark-x2.6-4b.json b/data/models/spark-x2.6-4b.json
new file mode 100644
index 00000000..0037c0fb
--- /dev/null
+++ b/data/models/spark-x2.6-4b.json
@@ -0,0 +1,69 @@
+{
+  "id": "spark-x2.6-4b",
+  "object": "model",
+  "created": 1792540800,
+  "owned_by": "iflytek",
+  "name": "Spark-X2.6-4B",
+  "api": {
+    "types": [
+      "responses"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": 1048576,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": []
+  },
+  "pricing": null,
+  "features": [],
+  "info": {
+    "status": "active",
+    "release_date": "2026-10-21",
+    "knowledge_cutoff": null,
+    "description": "iFlytek Spark-X2.6-4B. The MaaS catalog says code, agent, long-text, and multilingual ability are improved and that three reasoning modes were added. Context is tagged 1M, with text input and text output. It prints an OpenAI base at /v2, an Anthropic base at /anthropic, and a complete Responses URL at /v1/responses; only that Responses path is recorded. The catalog marks the model limited-time free and lists a zero inference price, so the standard price is unset.",
+    "docs": "https://xinghuo.xfyun.cn/sparkapi",
+    "verified_at": "2026-10-24"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "讯飞 Spark-X2.6-4B。星辰 MaaS 目录写明代码、智能体、长文本和多语言能力提升，并新增三档推理模式。上下文标为 1M，输入和输出都是文本。目录给出 OpenAI 基址 /v2、Anthropic 基址 /anthropic，以及完整的 Responses 地址 /v1/responses；这里只登记这条完整路径。目录角标为限时免费，推理价格为 0，因此标准价留空。"
+    }
+  },
+  "sources": [
+    {
+      "title": "讯飞星火大模型 API",
+      "url": "https://xinghuo.xfyun.cn/sparkapi",
+      "retrieved_at": "2026-10-24"
+    },
+    {
+      "title": "星辰 MaaS 模型目录（Spark-X2.6-4B）",
+      "url": "https://maas.xfyun.cn/api/v1/gpt-finetune/model/base/list-v2",
+      "retrieved_at": "2026-10-24"
+    }
+  ],
+  "providers": [
+    {
+      "id": "iflytek",
+      "model": "spark-x2.6-4b",
+      "base_url": "https://maas-api.cn-huabei-1.xf-yun.com",
+      "endpoints": {
+        "responses": "POST /v1/responses"
+      },
+      "reasoning_param": null
+    }
+  ]
+}
diff --git a/data/models/spark-x2.6.json b/data/models/spark-x2.6.json
new file mode 100644
index 00000000..1ffdeb7c
--- /dev/null
+++ b/data/models/spark-x2.6.json
@@ -0,0 +1,69 @@
+{
+  "id": "spark-x2.6",
+  "object": "model",
+  "created": 1792108800,
+  "owned_by": "iflytek",
+  "name": "Spark-X2.6",
+  "api": {
+    "types": [
+      "responses"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": 1048576,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": null,
+    "efforts": []
+  },
+  "pricing": null,
+  "features": [],
+  "info": {
+    "status": "active",
+    "release_date": "2026-10-16",
+    "knowledge_cutoff": null,
+    "description": "iFlytek Spark-X2.6 is a 293B-A30B mixture-of-experts model. The MaaS catalog tags the context length as 1M and lists text input and text output. It prints an OpenAI base at /v2, an Anthropic base at /anthropic, and a complete Responses URL at /v1/responses; only that Responses path is recorded. The public Spark API page publishes only a limited-time 60% discount, so the standard price is unset.",
+    "docs": "https://xinghuo.xfyun.cn/sparkapi",
+    "verified_at": "2026-10-24"
+  },
+  "aliases": [],
+  "translations": {
+    "zh": {
+      "description": "讯飞 Spark-X2.6 是 293B-A30B 混合专家模型。星辰 MaaS 目录把上下文标为 1M，输入和输出都是文本。目录给出 OpenAI 基址 /v2、Anthropic 基址 /anthropic，以及完整的 Responses 地址 /v1/responses；这里只登记这条完整路径。星火 API 页面只公布了限时 6 折，因此标准价留空。"
+    }
+  },
+  "sources": [
+    {
+      "title": "讯飞星火大模型 API",
+      "url": "https://xinghuo.xfyun.cn/sparkapi",
+      "retrieved_at": "2026-10-24"
+    },
+    {
+      "title": "星辰 MaaS 模型目录（Spark-X2.6）",
+      "url": "https://maas.xfyun.cn/api/v1/gpt-finetune/model/base/list-v2",
+      "retrieved_at": "2026-10-24"
+    }
+  ],
+  "providers": [
+    {
+      "id": "iflytek",
+      "model": "spark-x2.6",
+      "base_url": "https://maas-api.cn-huabei-1.xf-yun.com",
+      "endpoints": {
+        "responses": "POST /v1/responses"
+      },
+      "reasoning_param": null
+    }
+  ]
+}
diff --git a/data/models/voxtral-mini-transcribe-3-0.json b/data/models/voxtral-mini-transcribe-3-0.json
new file mode 100644
index 00000000..9e98be4b
--- /dev/null
+++ b/data/models/voxtral-mini-transcribe-3-0.json
@@ -0,0 +1,86 @@
+{
+  "id": "voxtral-mini-transcribe-3-0",
+  "object": "model",
+  "created": 1792454400,
+  "owned_by": "mistral",
+  "name": "Voxtral Mini Transcribe 3",
+  "api": {
+    "types": [
+      "audio"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": 32768,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "audio"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": null,
+    "efforts": []
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "minute",
+    "input": 0.0035,
+    "output": null,
+    "cache_read": 0.00035,
+    "cache_write": null
+  },
+  "features": [
+    "caching"
+  ],
+  "info": {
+    "status": "preview",
+    "release_date": "2026-10-20",
+    "knowledge_cutoff": null,
+    "description": "Public-preview transcription and diarization model. Context length is 32k. Input is audio and output is text.",
+    "docs": "https://docs.mistral.ai/models/voxtral-mini-transcribe-3-0",
+    "verified_at": "2026-10-22"
+  },
+  "aliases": [
+    "voxtral-mini-transcribe-3",
+    "voxtral-mini-transcribe-latest"
+  ],
+  "translations": {
+    "zh": {
+      "description": "公开预览的转写和说话人分离模型。上下文长度为 32k。输入为音频，输出为文本。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Voxtral Mini Transcribe 3 | Mistral Docs",
+      "url": "https://docs.mistral.ai/models/voxtral-mini-transcribe-3-0",
+      "retrieved_at": "2026-10-22"
+    },
+    {
+      "title": "API pricing | Mistral Docs",
+      "url": "https://docs.mistral.ai/inference/pricing",
+      "retrieved_at": "2026-10-22"
+    },
+    {
+      "title": "Migrate from Voxtral 2 | Mistral Docs",
+      "url": "https://docs.mistral.ai/studio/audio/migrate-from-voxtral-2",
+      "retrieved_at": "2026-10-22"
+    }
+  ],
+  "providers": [
+    {
+      "id": "mistral",
+      "model": "voxtral-mini-transcribe-3-0",
+      "base_url": "https://api.mistral.ai",
+      "endpoints": {
+        "audio": "POST /v1/audio/transcriptions"
+      },
+      "reasoning_param": null
+    }
+  ]
+}
diff --git a/data/models/voxtral-mini-transcribe-realtime-3-0.json b/data/models/voxtral-mini-transcribe-realtime-3-0.json
new file mode 100644
index 00000000..554e7a10
--- /dev/null
+++ b/data/models/voxtral-mini-transcribe-realtime-3-0.json
@@ -0,0 +1,94 @@
+{
+  "id": "voxtral-mini-transcribe-realtime-3-0",
+  "object": "model",
+  "created": 1792454400,
+  "owned_by": "mistral",
+  "name": "Voxtral Mini Transcribe Realtime 3",
+  "api": {
+    "types": [
+      "audio",
+      "realtime"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "audio"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": null,
+    "efforts": []
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "minute",
+    "input": 0.008,
+    "output": null,
+    "cache_read": 0.0008,
+    "cache_write": null
+  },
+  "features": [
+    "streaming",
+    "caching"
+  ],
+  "info": {
+    "status": "preview",
+    "release_date": "2026-10-20",
+    "knowledge_cutoff": null,
+    "description": "Public-preview realtime transcription model for a single WebSocket session, with configurable latency from 80 ms to 2,400 ms, end-of-turn detection, context biasing, and speaker diarization. Input is audio and output is text. The model card does not list a context length.",
+    "docs": "https://docs.mistral.ai/models/voxtral-mini-transcribe-realtime-3-0",
+    "verified_at": "2026-10-22"
+  },
+  "aliases": [
+    "voxtral-mini-transcribe-realtime-3",
+    "voxtral-mini-transcribe-realtime-latest"
+  ],
+  "translations": {
+    "zh": {
+      "description": "公开预览的实时转写模型，在单条 WebSocket 会话中提供 80 ms 到 2,400 ms 的可配置延迟、话轮检测、上下文偏置和说话人分离。输入为音频，输出为文本。模型卡未列出上下文长度。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Voxtral Mini Transcribe Realtime 3 | Mistral Docs",
+      "url": "https://docs.mistral.ai/models/voxtral-mini-transcribe-realtime-3-0",
+      "retrieved_at": "2026-10-22"
+    },
+    {
+      "title": "Realtime transcription | Mistral Docs",
+      "url": "https://docs.mistral.ai/studio/audio/speech-to-text/transcribe-in-realtime",
+      "retrieved_at": "2026-10-22"
+    },
+    {
+      "title": "API pricing | Mistral Docs",
+      "url": "https://docs.mistral.ai/inference/pricing",
+      "retrieved_at": "2026-10-22"
+    },
+    {
+      "title": "Migrate from Voxtral 2 | Mistral Docs",
+      "url": "https://docs.mistral.ai/studio/audio/migrate-from-voxtral-2",
+      "retrieved_at": "2026-10-22"
+    }
+  ],
+  "providers": [
+    {
+      "id": "mistral",
+      "model": "voxtral-mini-transcribe-realtime-3-0",
+      "base_url": "wss://api.mistral.ai",
+      "endpoints": {
+        "audio": "WS /v1/audio/transcriptions/realtime",
+        "realtime": "WS /v1/audio/transcriptions/realtime"
+      },
+      "reasoning_param": null
+    }
+  ]
+}
diff --git a/data/models/voxtral-mini-tts-3-0.json b/data/models/voxtral-mini-tts-3-0.json
new file mode 100644
index 00000000..0c391374
--- /dev/null
+++ b/data/models/voxtral-mini-tts-3-0.json
@@ -0,0 +1,90 @@
+{
+  "id": "voxtral-mini-tts-3-0",
+  "object": "model",
+  "created": 1792454400,
+  "owned_by": "mistral",
+  "name": "Voxtral TTS 3",
+  "api": {
+    "types": [
+      "audio"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text",
+      "audio"
+    ],
+    "output": [
+      "audio"
+    ]
+  },
+  "reasoning": {
+    "supported": null,
+    "efforts": []
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "1M_characters",
+    "input": 13,
+    "output": null,
+    "cache_read": 1.3,
+    "cache_write": null
+  },
+  "features": [
+    "streaming",
+    "caching"
+  ],
+  "info": {
+    "status": "preview",
+    "release_date": "2026-10-20",
+    "knowledge_cutoff": null,
+    "description": "Public-preview text-to-speech model with zero-shot and professional voice cloning, 9 languages, and streaming with about 90 ms time-to-first-audio. Input is text and audio; output is audio.",
+    "docs": "https://docs.mistral.ai/models/voxtral-mini-tts-3-0",
+    "verified_at": "2026-10-22"
+  },
+  "aliases": [
+    "voxtral-mini-tts-3",
+    "voxtral-mini-tts-realtime-3-0",
+    "voxtral-mini-tts-realtime-3",
+    "voxtral-mini-tts-realtime-latest"
+  ],
+  "translations": {
+    "zh": {
+      "description": "公开预览的文字转语音模型，支持零样本和专业声音克隆、9 种语言，流式输出的首个音频延迟约 90 ms。输入为文本和音频，输出为音频。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Voxtral TTS 3 | Mistral Docs",
+      "url": "https://docs.mistral.ai/models/voxtral-mini-tts-3-0",
+      "retrieved_at": "2026-10-22"
+    },
+    {
+      "title": "API pricing | Mistral Docs",
+      "url": "https://docs.mistral.ai/inference/pricing",
+      "retrieved_at": "2026-10-22"
+    },
+    {
+      "title": "Migrate from Voxtral 2 | Mistral Docs",
+      "url": "https://docs.mistral.ai/studio/audio/migrate-from-voxtral-2",
+      "retrieved_at": "2026-10-22"
+    }
+  ],
+  "providers": [
+    {
+      "id": "mistral",
+      "model": "voxtral-mini-tts-3-0",
+      "base_url": "https://api.mistral.ai",
+      "endpoints": {
+        "audio": "POST /v1/audio/speech"
+      },
+      "reasoning_param": null
+    }
+  ]
+}
diff --git a/data/models/voxtral-small-transcribe-3-0.json b/data/models/voxtral-small-transcribe-3-0.json
new file mode 100644
index 00000000..3e27bd9f
--- /dev/null
+++ b/data/models/voxtral-small-transcribe-3-0.json
@@ -0,0 +1,87 @@
+{
+  "id": "voxtral-small-transcribe-3-0",
+  "object": "model",
+  "created": 1792454400,
+  "owned_by": "mistral",
+  "name": "Voxtral Small Transcribe 3",
+  "api": {
+    "types": [
+      "audio"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": 32768,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "audio"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": null,
+    "efforts": []
+  },
+  "pricing": {
+    "currency": "USD",
+    "unit": "minute",
+    "input": 0.004,
+    "output": null,
+    "cache_read": 0.0004,
+    "cache_write": null
+  },
+  "features": [
+    "tools",
+    "caching"
+  ],
+  "info": {
+    "status": "preview",
+    "release_date": "2026-10-20",
+    "knowledge_cutoff": null,
+    "description": "Public-preview flagship transcription and diarization model. It also supports translation, speech question answering, summarization, and function calling. Context length is 32k. Input is audio and output is text.",
+    "docs": "https://docs.mistral.ai/models/voxtral-small-transcribe-3-0",
+    "verified_at": "2026-10-22"
+  },
+  "aliases": [
+    "voxtral-small-transcribe-3",
+    "voxtral-small-transcribe-latest"
+  ],
+  "translations": {
+    "zh": {
+      "description": "公开预览的旗舰转写和说话人分离模型。同时支持翻译、语音问答、摘要和函数调用。上下文长度为 32k。输入为音频，输出为文本。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Voxtral Small Transcribe 3 | Mistral Docs",
+      "url": "https://docs.mistral.ai/models/voxtral-small-transcribe-3-0",
+      "retrieved_at": "2026-10-22"
+    },
+    {
+      "title": "API pricing | Mistral Docs",
+      "url": "https://docs.mistral.ai/inference/pricing",
+      "retrieved_at": "2026-10-22"
+    },
+    {
+      "title": "Migrate from Voxtral 2 | Mistral Docs",
+      "url": "https://docs.mistral.ai/studio/audio/migrate-from-voxtral-2",
+      "retrieved_at": "2026-10-22"
+    }
+  ],
+  "providers": [
+    {
+      "id": "mistral",
+      "model": "voxtral-small-transcribe-3-0",
+      "base_url": "https://api.mistral.ai",
+      "endpoints": {
+        "audio": "POST /v1/audio/transcriptions"
+      },
+      "reasoning_param": null
+    }
+  ]
+}
diff --git a/data/models/xing4.0-29b-a4b.json b/data/models/xing4.0-29b-a4b.json
new file mode 100644
index 00000000..3ad40e05
--- /dev/null
+++ b/data/models/xing4.0-29b-a4b.json
@@ -0,0 +1,58 @@
+{
+  "id": "xing4.0-29b-a4b",
+  "object": "model",
+  "created": 1789603200,
+  "owned_by": "teleai",
+  "name": "Xing4.0-29B-A4B",
+  "api": {
+    "types": [
+      "chat"
+    ],
+    "sdk": []
+  },
+  "limits": {
+    "context": null,
+    "input": null,
+    "output": null
+  },
+  "modalities": {
+    "input": [
+      "text"
+    ],
+    "output": [
+      "text"
+    ]
+  },
+  "reasoning": {
+    "supported": true,
+    "efforts": []
+  },
+  "pricing": null,
+  "features": [
+    "tools"
+  ],
+  "info": {
+    "status": "active",
+    "release_date": "2026-09-17",
+    "knowledge_cutoff": null,
+    "description": "Open-weight model from China Telecom Artificial Intelligence Technology Co., Ltd., with 29B total parameters and 4B activated per token. The README says it natively supports a 256K context length, extensible to 512K, and the serving examples pass 262144 as a launch flag. TeleAI is not in the abbreviation table and the page does not state one exact context limit, so the context field is unset. It supports tool calling, and the sample output contains a chain of thought. No standard API token price or first-party request URL is published.",
+    "docs": "https://github.com/XingChen-AGI/Xing4.0-29B-A4B",
+    "verified_at": "2026-10-27"
+  },
+  "aliases": [
+    "Xing4.0-29B-A4B"
+  ],
+  "translations": {
+    "zh": {
+      "description": "中国电信人工智能科技有限公司的开放权重模型，总参数 29B，每 token 激活 4B。README 写原生支持 256K 上下文，可扩展到 512K；服务示例把 262144 作为启动参数。TeleAI 不在缩写换算表中，页面也没有给出唯一的精确上下文上限，因此上下文字段留空。它支持工具调用，示例输出包含思维链。没有公布 API 标准 token 价格或第一方请求 URL。"
+    }
+  },
+  "sources": [
+    {
+      "title": "Xing4.0-29B-A4B | XingChen-AGI",
+      "url": "https://github.com/XingChen-AGI/Xing4.0-29B-A4B",
+      "retrieved_at": "2026-10-27"
+    }
+  ],
+  "providers": []
+}
diff --git a/data/vendors/aionlabs.json b/data/vendors/aionlabs.json
new file mode 100644
index 00000000..5cf43072
--- /dev/null
+++ b/data/vendors/aionlabs.json
@@ -0,0 +1,5 @@
+{
+  "id": "aionlabs",
+  "name": "Aion Labs",
+  "url": "https://www.aionlabs.ai"
+}
diff --git a/data/vendors/aleph-alpha.json b/data/vendors/aleph-alpha.json
new file mode 100644
index 00000000..346ed4d8
--- /dev/null
+++ b/data/vendors/aleph-alpha.json
@@ -0,0 +1,5 @@
+{
+  "id": "aleph-alpha",
+  "name": "Aleph Alpha",
+  "url": "https://huggingface.co/Aleph-Alpha"
+}
diff --git a/data/vendors/elorian.json b/data/vendors/elorian.json
new file mode 100644
index 00000000..3914d561
--- /dev/null
+++ b/data/vendors/elorian.json
@@ -0,0 +1,5 @@
+{
+  "id": "elorian",
+  "name": "Elorian",
+  "url": "https://elorian.ai"
+}
diff --git a/data/vendors/fireworks.json b/data/vendors/fireworks.json
new file mode 100644
index 00000000..4f8ed6b1
--- /dev/null
+++ b/data/vendors/fireworks.json
@@ -0,0 +1,5 @@
+{
+  "id": "fireworks",
+  "name": "Fireworks",
+  "url": "https://fireworks.ai"
+}
diff --git a/data/vendors/huggingface-bio.json b/data/vendors/huggingface-bio.json
new file mode 100644
index 00000000..b503dfd0
--- /dev/null
+++ b/data/vendors/huggingface-bio.json
@@ -0,0 +1,5 @@
+{
+  "id": "huggingface-bio",
+  "name": "Hugging Face Bio",
+  "url": "https://huggingface.co/HuggingFaceBio"
+}
diff --git a/data/vendors/inclusionai.json b/data/vendors/inclusionai.json
new file mode 100644
index 00000000..da1d50b0
--- /dev/null
+++ b/data/vendors/inclusionai.json
@@ -0,0 +1,5 @@
+{
+  "id": "inclusionai",
+  "name": "inclusionAI",
+  "url": "https://huggingface.co/inclusionAI"
+}
diff --git a/data/vendors/liquid.json b/data/vendors/liquid.json
new file mode 100644
index 00000000..172f3db7
--- /dev/null
+++ b/data/vendors/liquid.json
@@ -0,0 +1,5 @@
+{
+  "id": "liquid",
+  "name": "Liquid AI",
+  "url": "https://www.liquid.ai"
+}
diff --git a/data/vendors/moondream.json b/data/vendors/moondream.json
new file mode 100644
index 00000000..42993891
--- /dev/null
+++ b/data/vendors/moondream.json
@@ -0,0 +1,5 @@
+{
+  "id": "moondream",
+  "name": "Moondream",
+  "url": "https://moondream.ai"
+}
diff --git a/data/vendors/openbmb.json b/data/vendors/openbmb.json
new file mode 100644
index 00000000..fd85078a
--- /dev/null
+++ b/data/vendors/openbmb.json
@@ -0,0 +1,5 @@
+{
+  "id": "openbmb",
+  "name": "OpenBMB",
+  "url": "https://openbmb.cn"
+}
diff --git a/data/vendors/perceptron.json b/data/vendors/perceptron.json
new file mode 100644
index 00000000..1fdebb7f
--- /dev/null
+++ b/data/vendors/perceptron.json
@@ -0,0 +1,5 @@
+{
+  "id": "perceptron",
+  "name": "Perceptron",
+  "url": "https://www.perceptron.inc"
+}
diff --git a/data/vendors/recursive.json b/data/vendors/recursive.json
new file mode 100644
index 00000000..893ca3bd
--- /dev/null
+++ b/data/vendors/recursive.json
@@ -0,0 +1,5 @@
+{
+  "id": "recursive",
+  "name": "Recursive",
+  "url": "https://recursive.com"
+}
diff --git a/data/vendors/reflection.json b/data/vendors/reflection.json
new file mode 100644
index 00000000..9c4c41b0
--- /dev/null
+++ b/data/vendors/reflection.json
@@ -0,0 +1,5 @@
+{
+  "id": "reflection",
+  "name": "Reflection",
+  "url": "https://reflection.ai"
+}
diff --git a/data/vendors/teleai.json b/data/vendors/teleai.json
new file mode 100644
index 00000000..63a00400
--- /dev/null
+++ b/data/vendors/teleai.json
@@ -0,0 +1,5 @@
+{
+  "id": "teleai",
+  "name": "TeleAI",
+  "url": "https://github.com/XingChen-AGI"
+}
diff --git a/data/vendors/unbiased.json b/data/vendors/unbiased.json
new file mode 100644
index 00000000..8be33ca8
--- /dev/null
+++ b/data/vendors/unbiased.json
@@ -0,0 +1,5 @@
+{
+  "id": "unbiased",
+  "name": "Unbiased AI",
+  "url": "https://unbiased.ai"
+}