feat(compute): refresh provider model catalog and compatibility profiles (#111)

__tests__/validate.test.mjs

@@ -251,7 +251,7 @@ describe('真实数据全量校验', () => {
       assert.ok(model.serviceType.includes('reasoning'))
       assert.equal(model.extra.supportsThinking, true)
       assert.equal(model.extra.thinkingDefault, true)
-      assert.deepEqual(model.extra.reasoningEffort, ['high', 'max'])
+      assert.deepEqual(model.extra.reasoning.supportedEfforts, ['low', 'high', 'max'])
 
       const specs = specFile.specs.filter((item) => item.id === modelId)
       assert.equal(specs.length, 1, `model-specs 中 ${modelId} 应且仅应有一条规格`)
@@ -264,7 +264,7 @@ describe('真实数据全量校验', () => {
         assert.ok(spec.serviceType.includes('reasoning'))
         assert.equal(spec.extra.supportsThinking, true)
         assert.equal(spec.extra.thinkingDefault, true)
-        assert.deepEqual(spec.extra.reasoningEffort, ['high', 'max'])
+        assert.equal('reasoning' in spec.extra, false)
       }
     }
   })
@@ -320,7 +320,7 @@ describe('真实数据全量校验', () => {
     assert.equal(model.extra.cacheHitPrice, 0.02)
     assert.equal(model.extra.concurrencyLimit, 2500)
     assert.deepEqual(model.extra.reasoning, {
-      supportedEfforts: ['high', 'max'],
+      supportedEfforts: ['low', 'high', 'max'],
       defaultEffort: 'high',
     })
     assert.deepEqual(model.extra.pricingTiers, [
@@ -329,7 +329,7 @@ describe('真实数据全量校验', () => {
     ])
 
     assert.ok(modelSpec, `model-specs 缺少 ${modelId}`)
-    assert.deepEqual(modelSpec.match.exact, [modelId, aliasId])
+    assert.deepEqual(modelSpec.match.exact, [modelId, aliasId, 'deepseek/deepseek-v4.1-flash'])
     assert.equal(
       specFile.specs.filter((item) => item.match?.exact?.includes(aliasId)).length,
       1,
@@ -340,7 +340,7 @@ describe('真实数据全量校验', () => {
     assert.ok(modelSpec.spec.capabilities.includes('vision'))
     assert.deepEqual(modelSpec.spec.serviceType, ['chat', 'reasoning', 'vision'])
     assert.equal(modelSpec.spec.supportsReasoning, true)
-    assert.deepEqual(modelSpec.routing.reasoning.supportedModes, ['auto', 'high', 'max'])
+    assert.deepEqual(modelSpec.routing.reasoning.supportedModes, ['auto', 'low', 'high', 'max'])
   })
 
   it('MiMo V2.5 ASR 应有独立精确规格，避免回落到 MiMo V2.5 family', () => {
@@ -647,7 +647,7 @@ describe('真实数据全量校验', () => {
       }
     }
     const gpt = read('model-specs', 'openai').specs.find((item) => item.id === 'gpt-6.1-sol')
-    assert.deepEqual(gpt.match.exact, ['gpt-6.1-sol'])
+    assert.deepEqual(gpt.match.exact, ['gpt-6.1-sol', 'openai/gpt-6.1-sol'])
     assert.deepEqual(gpt.routing.reasoning.supportedModes, ['auto', ...efforts])
     for (const provider of ['anthropic', 'anthropic-claude']) {
       const model = read('providers', provider).models.find((item) => item.modelName === 'claude-sonnet-5-5')
@@ -658,7 +658,7 @@ describe('真实数据全量校验', () => {
       if (provider === 'anthropic') assert.equal(model.extra.promptCacheMinTokens, 512)
     }
     const sonnet = read('model-specs', 'anthropic').specs.find((item) => item.id === 'claude-sonnet-5-5')
-    assert.deepEqual(sonnet.match.exact, ['claude-sonnet-5-5'])
+    assert.deepEqual(sonnet.match.exact, ['claude-sonnet-5-5', 'anthropic/claude-sonnet-5.5'])
     assert.equal(sonnet.spec.extra.forcedToolChoiceUnsupported, true)
     assert.equal(sonnet.spec.extra.thinkingOnly, undefined)
     assert.equal(sonnet.routing.reasoning.defaultMode, 'high')
@@ -676,10 +676,12 @@ describe('真实数据全量校验', () => {
     }
   })
 
-  it('GLM-5.3-FlashX 仅加入共享规格，避免假定中国区智谱 API 已提供该 ID', () => {
+  it('GLM-5.3-FlashX 官方中国区已开放 API，但套餐仍不提供', () => {
     const provider = JSON.parse(readFileSync(join(ROOT, 'compute', 'providers', 'zhipu.json'), 'utf8'))
     const specs = JSON.parse(readFileSync(join(ROOT, 'compute', 'model-specs', 'zhipu.json'), 'utf8')).specs
-    assert.equal(provider.models.some((item) => item.modelName === 'glm-5.3-flashx'), false)
+    assert.equal(provider.models.some((item) => item.modelName === 'glm-5.3-flashx'), true)
+    const plan = JSON.parse(readFileSync(join(ROOT, 'compute', 'coding-plans', 'zhipu-coding.json'), 'utf8'))
+    assert.equal(plan.models.some((item) => item.modelName === 'glm-5.3-flashx'), false)
     const spec = specs.find((item) => item.id === 'glm-5.3-flashx')
     assert.ok(spec)
     assert.deepEqual(spec.match.exact, ['glm-5.3-flashx', 'z-ai/glm-5.3-flashx'])
@@ -796,7 +798,7 @@ describe('真实数据全量校验', () => {
     // Anthropic Messages 中转只能从 ModelSpec 补全拿到声明，预置 Provider 条目不经补全，
     // 两处漏写都会回退成 budget_tokens（desirecore#2940）。Opus 4.6 / Sonnet 4.6 仍接受
     // budget_tokens（仅弃用），不在强制范围内。新代际若同样拒绝 budget_tokens，先扩展这里的正则。
-    const adaptiveGeneration = /^claude-(?:fable|mythos)(?:$|-)|^claude-(?:opus|sonnet)-(?:[5-9]|[1-9]\d)(?:$|-)|^claude-opus-4-[78](?:$|-)/
+    const adaptiveGeneration = /^claude-(?:fable|mythos)(?:$|-)|^claude-(?:opus|sonnet)-(?:[5-9]|[1-9]\d)(?:$|-)|^claude-opus-4-[78](?:$|-)|^minimax-m3-1-flash-preview$/
     const thinkingOnlyGeneration = /^claude-(?:fable|mythos)(?:$|-)|^claude-opus-5-5(?:$|-)/
     // 与客户端匹配器同口径：小写、去 vendor 前缀、统一分隔符
     const normalizeId = (id) => id.toLowerCase().trim().split('/').at(-1).replace(/[._:\s]+/g, '-')
@@ -917,6 +919,80 @@ describe('真实数据全量校验', () => {
     }
   })
 
+  it('全供应商更新保持官方接入边界、强制推理合同与非 token 计价', () => {
+    const read = (folder, name) => JSON.parse(readFileSync(join(ROOT, 'compute', folder, `${name}.json`), 'utf8'))
+    const findModel = (folder, name, id) => {
+      const row = read(folder, name).models.find((item) => item.modelName === id)
+      assert.ok(row, `${name} 缺少 ${id}`)
+      return row
+    }
+    const glm = findModel('providers', 'zhipu', 'glm-5.3')
+    assert.deepEqual(glm.extra.reasoning, { supportedEfforts: ['low', 'high', 'max'], defaultEffort: 'max' })
+    assert.equal(glm.inputPrice, 8)
+    assert.equal(glm.outputPrice, 28)
+    const flashx = findModel('providers', 'zhipu', 'glm-5.3-flashx')
+    assert.equal(flashx.inputPrice, 2)
+    assert.equal(flashx.outputPrice, 7)
+    const kimi = findModel('providers', 'moonshot', 'kimi-k3')
+    const kimiSpec = read('model-specs', 'moonshot').specs.find((item) => item.id === 'kimi-k3')
+    assert.equal(kimi.extra.thinking.disableSupported, false)
+    assert.equal(kimi.extra.samplingParametersDeprecated, true)
+    assert.equal(kimiSpec.spec.extra.samplingParametersDeprecated, true)
+    assert.equal(kimi.extra.reasoning.defaultEffort, 'max')
+    assert.deepEqual(kimiSpec.routing.reasoning.supportedModes, ['auto', 'low', 'high', 'max'])
+    assert.equal(kimi.maxOutputTokens, 1048576)
+    assert.equal(kimi.inputPrice, undefined, '不把国际美元价混写到国内人民币 Provider')
+    const previewId = 'MiniMax-M3.1-Flash-Preview'
+    const preview = findModel('coding-plans', 'minimax-coding', previewId)
+    assert.equal(preview.extra.thinkingOnly, true)
+    assert.equal(preview.extra.reasoning.defaultEffort, 'max')
+    assert.equal(read('providers', 'minimax').models.some((item) => item.modelName === previewId), false)
+    const embed = findModel('providers', 'cohere', 'embed-v5.0-pro')
+    assert.deepEqual(embed.extra.dimensions, [256, 512, 768, 1024, 1536, 2048])
+    assert.equal(embed.inputPrice, undefined)
+    assert.equal(read('providers', 'cohere').models.some((item) => item.modelName === 'rerank-v4.0-fast'), false, '不向 OpenAI 兼容端点声明原生 Rerank 接入')
+    assert.ok(read('model-specs', 'cohere').specs.some((item) => item.id === 'rerank-v4.0-fast'))
+    const command = findModel('providers', 'cohere', 'command-a-plus-05-2026')
+    assert.deepEqual(command.extra.reasoning.supportedEfforts, ['none', 'high'])
+    assert.equal(embed.capabilities.includes('vision'), false, '兼容 Embeddings 接入仅接受文本')
+    const small = findModel('providers', 'mistral', 'mistral-small-latest')
+    assert.equal(small.contextWindow, 262144)
+    assert.equal(small.inputPrice, 0.15)
+    assert.equal(small.outputPrice, 0.6)
+    const omni = findModel('providers', 'dashscope', 'qwen3.8-omni-flash')
+    assert.equal(omni.inputPrice, 0.8)
+    assert.equal(omni.outputPrice, 2.7)
+    assert.deepEqual(omni.serviceType, ['chat', 'vision'])
+    assert.equal(omni.capabilities.includes('tts'), false, 'Omni Flash 当前只输出文本')
+    assert.equal(findModel('providers', 'google', 'gemini-embedding-2').contextWindow, 8192)
+    assert.ok(findModel('providers', 'kling', 'kling-v3'))
+    assert.equal(read('providers', 'kling').models.some((item) => item.modelName === 'kling-3.0-turbo'), false)
+    assert.equal(read('providers', 'tencent').models.some((item) => item.modelName === 'hy4-preview'), false)
+    for (const id of ['deepseek-v4.1-flash', 'glm-5.3', 'qwen-image-3.0-pro', 'happyhorse-1.1-t2v']) {
+      assert.ok(findModel('coding-plans', 'dashscope-token-plan', id))
+    }
+  })
+
+  it('新版 Mistral 与 Command A 规格的 family 不误吞其他代际或子型号', () => {
+    const readSpecs = (name) => JSON.parse(readFileSync(join(ROOT, 'compute', 'model-specs', `${name}.json`), 'utf8')).specs
+    const normalize = (id) => id.toLowerCase().split('/').at(-1).replace(/[._]/g, '-')
+    const keys = (row) => [row.id, row.family, ...(row.match?.exact ?? []), ...(row.match?.patterns ?? []).map((p) => p.split('*')[0])].filter(Boolean).map(normalize)
+    for (const [id, forbidden] of [
+      ['mistral-medium-3.5', ['mistral-medium-2508', 'mistral-medium-2312']],
+      ['mistral-small-latest', ['mistral-small-2506', 'mistral-small-2402']],
+      ['mistral-large-latest', ['mistral-large-2411', 'mistral-large-2407']],
+    ]) {
+      const row = readSpecs('mistral').find((item) => item.id === id)
+      for (const oldId of forbidden) {
+        assert.equal(keys(row).some((key) => normalize(oldId) === key || normalize(oldId).startsWith(key + '-')), false, `${id} 的匹配键会误吞 ${oldId}`)
+      }
+    }
+    const command = readSpecs('cohere').find((item) => item.id === 'command-a-03-2025')
+    assert.equal(normalize(command.family) === 'command-a' || 'command-a-plus'.startsWith(normalize(command.family) + '-'), false)
+    assert.equal((command.match.patterns ?? []).length, 0)
+    assert.equal((command.match.exact ?? []).map(normalize).includes('command-a-plus'), false)
+  })
+
   it('所有 Provider 应按供应商归属计价，模型来源不覆盖供应商币种', () => {
     const expectedCurrencies = {
       anthropic: 'USD',

compute/coding-plans/dashscope-token-plan.json

@@ -27,6 +27,297 @@
     }
   },
   "models": [
+    {
+      "modelName": "deepseek-v4.1-flash",
+      "displayName": "DeepSeek V4.1 Flash",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 384000,
+      "capabilities": [
+        "chat",
+        "code",
+        "reasoning",
+        "deep_thinking",
+        "vision",
+        "multilingual",
+        "tool_use"
+      ],
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "extra": {
+        "supportsThinking": true,
+        "thinkingDefault": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "high"
+        }
+      },
+      "description": "DeepSeek V4.1 Flash，支持文本与图像输入，100 万上下文",
+      "source": "preset"
+    },
+    {
+      "modelName": "wan2.7-image-pro",
+      "displayName": "通义万相 2.7 Pro",
+      "capabilities": [
+        "image_generation",
+        "chinese_optimized",
+        "high_quality"
+      ],
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "万相最新旗舰文生图模型，支持 4K 分辨率",
+      "source": "preset",
+      "extra": {}
+    },
+    {
+      "modelName": "wan2.7-image",
+      "displayName": "通义万相 2.7",
+      "capabilities": [
+        "image_generation",
+        "chinese_optimized"
+      ],
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "万相最新标准文生图模型，高画质",
+      "source": "preset",
+      "extra": {}
+    },
+    {
+      "modelName": "qwen-image-3.0-pro",
+      "displayName": "qwen-image-3.0-pro",
+      "capabilities": [
+        "image_generation",
+        "image_editing",
+        "text_rendering"
+      ],
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "qwen-image-3.0-pro",
+      "source": "preset"
+    },
+    {
+      "modelName": "happyhorse-1.1-t2v",
+      "displayName": "HappyHorse 1.1 T2V",
+      "capabilities": [
+        "video_generation",
+        "text_to_video",
+        "high_quality"
+      ],
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "HappyHorse 1.1 文生视频，进一步提升文本词义理解、镜头调度与动态生成表现",
+      "source": "preset",
+      "extra": {}
+    },
+    {
+      "modelName": "happyhorse-1.1-r2v",
+      "displayName": "HappyHorse 1.1 R2V",
+      "capabilities": [
+        "video_generation",
+        "reference_to_video",
+        "multi_image",
+        "high_quality"
+      ],
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "HappyHorse 1.1 参考生视频，支持最多9张图片参考，更加稳定的主体与场景参考",
+      "source": "preset",
+      "extra": {}
+    },
+    {
+      "modelName": "happyhorse-1.1-i2v",
+      "displayName": "HappyHorse 1.1 I2V",
+      "capabilities": [
+        "video_generation",
+        "image_to_video",
+        "high_quality"
+      ],
+      "serviceType": [
+        "video_gen"
+      ],
+      "description": "HappyHorse 1.1 图生视频，进一步提升画面质感、动态表现与跨片段一致性",
+      "source": "preset",
+      "extra": {}
+    },
+    {
+      "modelName": "glm-5.3",
+      "displayName": "GLM-5.3",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "math",
+        "multilingual",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "serviceType": [
+        "chat"
+      ],
+      "description": "智谱 GLM-5.3 大推理模型，面向复杂软件工程与长程 Agent 任务，100 万上下文；强制思考，无法关闭",
+      "extra": {
+        "thinkingDefault": true,
+        "thinkingOnly": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "max"
+        }
+      },
+      "source": "preset"
+    },
+    {
+      "modelName": "glm-5.2",
+      "displayName": "GLM-5.2",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 32768,
+      "capabilities": [
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "multilingual",
+        "tool_use",
+        "long_context"
+      ],
+      "serviceType": [
+        "chat"
+      ],
+      "description": "智谱 GLM-5.2 大推理模型，100 万上下文",
+      "source": "preset",
+      "extra": {
+        "reasoning": {
+          "supportedEfforts": [
+            "minimal",
+            "low",
+            "medium",
+            "high",
+            "xhigh"
+          ],
+          "defaultEffort": "high"
+        }
+      }
+    },
+    {
+      "modelName": "deepseek-v4-pro-0813",
+      "displayName": "DeepSeek V4 Pro 0813",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 384000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "math",
+        "science",
+        "multilingual",
+        "tool_use"
+      ],
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "extra": {
+        "supportsThinking": true,
+        "thinkingDefault": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "high"
+        }
+      },
+      "description": "DeepSeek V4 Pro 0813 版本，1.6T MoE 架构，100 万上下文",
+      "source": "preset"
+    },
+    {
+      "modelName": "deepseek-v4-pro",
+      "displayName": "DeepSeek V4 Pro",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 384000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "math",
+        "science",
+        "multilingual",
+        "tool_use"
+      ],
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "extra": {
+        "supportsThinking": true,
+        "thinkingDefault": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "high"
+        }
+      },
+      "description": "DeepSeek V4 Pro，1.6T MoE 架构，100 万上下文",
+      "source": "preset"
+    },
+    {
+      "modelName": "deepseek-v4-flash-0731",
+      "displayName": "DeepSeek V4 Flash 0731",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 384000,
+      "capabilities": [
+        "chat",
+        "code",
+        "reasoning",
+        "deep_thinking",
+        "multilingual",
+        "tool_use"
+      ],
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "extra": {
+        "supportsThinking": true,
+        "thinkingDefault": true,
+        "reasoningEffort": [
+          "high",
+          "max"
+        ],
+        "reasoning": {
+          "supportedEfforts": [
+            "high",
+            "max"
+          ],
+          "defaultEffort": "high"
+        }
+      },
+      "description": "DeepSeek V4 Flash 0731 固定版本，100 万上下文",
+      "source": "preset"
+    },
     {
       "modelName": "qwen3.8-max",
       "displayName": "Qwen3.8 Max (Token Plan)",

compute/coding-plans/minimax-coding.json

@@ -29,6 +29,42 @@
     }
   },
   "models": [
+    {
+      "modelName": "MiniMax-M3.1-Flash-Preview",
+      "displayName": "MiniMax M3.1 Flash Preview",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "video_understanding",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "description": "MiniMax M3.1 Flash Preview，目前仅通过 M Plan / MiniMax Code 提供",
+      "contextWindow": 1000000,
+      "extra": {
+        "adaptiveThinking": true,
+        "thinkingOnly": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "medium",
+            "high",
+            "xhigh",
+            "max"
+          ],
+          "defaultEffort": "max"
+        }
+      },
+      "source": "preset"
+    },
     {
       "modelName": "MiniMax-M3",
       "displayName": "MiniMax M3 (Token Plan)",

compute/coding-plans/zhipu-coding.json

@@ -27,6 +27,43 @@
     "glm-4.5-air"
   ],
   "models": [
+    {
+      "modelName": "glm-5.3",
+      "displayName": "GLM-5.3",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "math",
+        "multilingual",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "serviceType": [
+        "chat"
+      ],
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "description": "智谱 GLM-5.3 大推理模型，面向复杂软件工程与长程 Agent 任务，100 万上下文；强制思考，无法关闭",
+      "extra": {
+        "thinkingDefault": true,
+        "thinkingOnly": true,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "max"
+        },
+        "pricingNotes": "国内官方标准价，人民币/百万 tokens；Flash 限时优惠以控制台实际结算为准。"
+      },
+      "source": "preset"
+    },
     {
       "modelName": "glm-5.3-flash",
       "displayName": "GLM-5.3-Flash (Coding)",

compute/model-specs/anthropic.json

@@ -7,7 +7,8 @@
       "family": "claude-opus-5-5",
       "match": {
         "exact": [
-          "claude-opus-5-5"
+          "claude-opus-5-5",
+          "anthropic/claude-opus-5.5"
         ]
       },
       "spec": {
@@ -242,7 +243,8 @@
       "family": "claude-sonnet-5-5",
       "match": {
         "exact": [
-          "claude-sonnet-5-5"
+          "claude-sonnet-5-5",
+          "anthropic/claude-sonnet-5.5"
         ]
       },
       "spec": {

compute/model-specs/cohere.json

@@ -1,6 +1,141 @@
 {
   "description": "Cohere 系列模型规格。参数来源：config-center compute/providers/cohere.json。包含对话、Embedding 和 Rerank 三类服务。",
   "specs": [
+    {
+      "id": "rerank-v4.0-fast",
+      "displayName": "Rerank 4 Fast",
+      "family": "rerank-v4.0-fast",
+      "match": {
+        "exact": [
+          "rerank-v4.0-fast"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "rerank",
+          "multilingual"
+        ],
+        "serviceType": [
+          "rerank"
+        ],
+        "description": "Rerank 4 Fast"
+      }
+    },
+    {
+      "id": "embed-v5.0-fast",
+      "displayName": "Embed 5 Fast",
+      "family": "embed-v5.0-fast",
+      "match": {
+        "exact": [
+          "embed-v5.0-fast"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "text_embedding",
+          "vision",
+          "multilingual"
+        ],
+        "serviceType": [
+          "embedding"
+        ],
+        "description": "Embed 5 Fast",
+        "contextWindow": 128000,
+        "extra": {
+          "dimensions": [
+            256,
+            512,
+            768,
+            1024,
+            1536,
+            2048
+          ],
+          "defaultDimension": 2048
+        },
+        "releasedAt": "2026-09-30"
+      }
+    },
+    {
+      "id": "rerank-v4.0-pro",
+      "displayName": "Rerank 4 Pro",
+      "family": "rerank-v4.0-pro",
+      "match": {
+        "exact": [
+          "rerank-v4.0-pro"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "rerank",
+          "multilingual"
+        ],
+        "serviceType": [
+          "rerank"
+        ],
+        "description": "Rerank 4 Pro"
+      }
+    },
+    {
+      "id": "embed-v5.0-pro",
+      "displayName": "Embed 5 Pro",
+      "family": "embed-v5.0-pro",
+      "match": {
+        "exact": [
+          "embed-v5.0-pro"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "text_embedding",
+          "vision",
+          "multilingual"
+        ],
+        "serviceType": [
+          "embedding"
+        ],
+        "description": "Embed 5 Pro",
+        "contextWindow": 128000,
+        "extra": {
+          "dimensions": [
+            256,
+            512,
+            768,
+            1024,
+            1536,
+            2048
+          ],
+          "defaultDimension": 2048
+        },
+        "releasedAt": "2026-09-30"
+      }
+    },
+    {
+      "id": "command-a-plus-05-2026",
+      "displayName": "Command A+",
+      "family": "command-a-plus-05-2026",
+      "match": {
+        "exact": [
+          "command-a-plus-05-2026"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "vision",
+          "tool_use",
+          "multilingual",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "description": "Command A+",
+        "contextWindow": 128000,
+        "maxOutputTokens": 64000,
+        "supportsReasoning": true
+      }
+    },
     {
       "id": "north-mini-code",
       "displayName": "North Mini Code",
@@ -31,13 +166,11 @@
     {
       "id": "command-a-03-2025",
       "displayName": "Command A",
-      "family": "command-a",
+      "family": "command-a-03-2025",
       "match": {
         "exact": [
-          "command-a-03-2025"
+          "command-a-03-2025",
-        ],
+          "cohere/command-a"
-        "patterns": [
-          "command-a*"
         ]
       },
       "spec": {
@@ -45,9 +178,7 @@
         "maxOutputTokens": 8000,
         "capabilities": [
           "chat",
-          "reasoning",
           "code",
-          "vision",
           "tool_use",
           "rag",
           "long_context"
@@ -56,7 +187,7 @@
           "chat"
         ],
         "defaultTemperature": 0.3,
-        "description": "Cohere 旗舰对话模型，256K 上下文，擅长 RAG 和工具调用"
+        "description": "Cohere Command A 文本模型，256K 上下文，支持 RAG 和工具调用"
       }
     },
     {

compute/model-specs/deepseek.json

@@ -31,11 +31,7 @@
         "supportsReasoning": true,
         "extra": {
           "supportsThinking": true,
-          "thinkingDefault": true,
+          "thinkingDefault": true
-          "reasoningEffort": [
-            "high",
-            "max"
-          ]
         },
         "description": "DeepSeek V4 Pro，1.6T MoE 架构，100 万上下文"
       },
@@ -47,6 +43,7 @@
         "reasoning": {
           "supportedModes": [
             "auto",
+            "low",
             "high",
             "max"
           ],
@@ -84,11 +81,7 @@
         "supportsReasoning": true,
         "extra": {
           "supportsThinking": true,
-          "thinkingDefault": true,
+          "thinkingDefault": true
-          "reasoningEffort": [
-            "high",
-            "max"
-          ]
         },
         "description": "DeepSeek V4 Pro 0813 版本，1.6T MoE 架构，100 万上下文"
       },
@@ -100,6 +93,7 @@
         "reasoning": {
           "supportedModes": [
             "auto",
+            "low",
             "high",
             "max"
           ],
@@ -166,7 +160,8 @@
       "match": {
         "exact": [
           "deepseek-flash",
-          "deepseek-v4.1-flash"
+          "deepseek-v4.1-flash",
+          "deepseek/deepseek-v4.1-flash"
         ]
       },
       "spec": {
@@ -202,6 +197,7 @@
         "reasoning": {
           "supportedModes": [
             "auto",
+            "low",
             "high",
             "max"
           ],
@@ -237,11 +233,7 @@
         "supportsReasoning": true,
         "extra": {
           "supportsThinking": true,
-          "thinkingDefault": true,
+          "thinkingDefault": true
-          "reasoningEffort": [
-            "high",
-            "max"
-          ]
         }
       },
       "routing": {
@@ -252,6 +244,7 @@
         "reasoning": {
           "supportedModes": [
             "auto",
+            "low",
             "high",
             "max"
           ],
@@ -341,11 +334,7 @@
         "releasedAt": "2026-08-21",
         "extra": {
           "supportsThinking": true,
-          "thinkingDefault": true,
+          "thinkingDefault": true
-          "reasoningEffort": [
-            "high",
-            "max"
-          ]
         },
         "description": "DeepSeek V4 Flash 实验性视觉版本，支持文本与图像输入，文本能力与 V4 Flash 一致"
       },
@@ -357,6 +346,7 @@
         "reasoning": {
           "supportedModes": [
             "auto",
+            "low",
             "high",
             "max"
           ],

compute/model-specs/google.json

@@ -1,6 +1,159 @@
 {
   "description": "Google Gemini 系列模型规格。参数来源：config-center compute/providers/google.json。",
   "specs": [
+    {
+      "id": "gemini-embedding-2",
+      "displayName": "Gemini Embedding 2",
+      "family": "gemini-embedding-2",
+      "match": {
+        "exact": [
+          "gemini-embedding-2"
+        ]
+      },
+      "spec": {
+        "contextWindow": 8192,
+        "capabilities": [
+          "text_embedding",
+          "vision",
+          "audio_understanding",
+          "video_understanding",
+          "multilingual"
+        ],
+        "serviceType": [
+          "embedding"
+        ],
+        "description": "Gemini Embedding 2，原生多模态嵌入，8192 token 输入，共享文本/图像/音视频向量空间",
+        "extra": {
+          "dimensions": [
+            768,
+            1536,
+            3072
+          ],
+          "defaultDimension": 3072
+        }
+      }
+    },
+    {
+      "id": "gemini-3.5-transcribe-live",
+      "displayName": "gemini-3.5-transcribe-live",
+      "family": "gemini-3.5-transcribe-live",
+      "match": {
+        "exact": [
+          "gemini-3.5-transcribe-live"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "asr",
+          "realtime"
+        ],
+        "serviceType": [
+          "asr"
+        ],
+        "description": "gemini-3.5-transcribe-live，官方模型目录；原生音频协议需客户端适配"
+      }
+    },
+    {
+      "id": "gemini-3.5-transcribe",
+      "displayName": "gemini-3.5-transcribe",
+      "family": "gemini-3.5-transcribe",
+      "match": {
+        "exact": [
+          "gemini-3.5-transcribe"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "asr"
+        ],
+        "serviceType": [
+          "asr"
+        ],
+        "description": "gemini-3.5-transcribe，官方模型目录；原生音频协议需客户端适配"
+      }
+    },
+    {
+      "id": "gemini-3.8-live-extended-thinking",
+      "displayName": "gemini-3.8-live-extended-thinking",
+      "family": "gemini-3.8-live-extended-thinking",
+      "match": {
+        "exact": [
+          "gemini-3.8-live-extended-thinking"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "realtime",
+          "speech_to_speech",
+          "reasoning"
+        ],
+        "serviceType": [
+          "realtime_voice"
+        ],
+        "description": "gemini-3.8-live-extended-thinking，官方模型目录；原生音频协议需客户端适配"
+      }
+    },
+    {
+      "id": "gemini-3.8-live",
+      "displayName": "gemini-3.8-live",
+      "family": "gemini-3.8-live",
+      "match": {
+        "exact": [
+          "gemini-3.8-live"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "realtime",
+          "speech_to_speech"
+        ],
+        "serviceType": [
+          "realtime_voice"
+        ],
+        "description": "gemini-3.8-live，官方模型目录；原生音频协议需客户端适配"
+      }
+    },
+    {
+      "id": "gemini-3.8-flash-lite-tts",
+      "displayName": "gemini-3.8-flash-lite-tts",
+      "family": "gemini-3.8-flash-lite-tts",
+      "match": {
+        "exact": [
+          "gemini-3.8-flash-lite-tts"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "tts",
+          "fast"
+        ],
+        "serviceType": [
+          "tts"
+        ],
+        "description": "gemini-3.8-flash-lite-tts，官方模型目录；原生音频协议需客户端适配"
+      }
+    },
+    {
+      "id": "gemini-3.8-flash-tts",
+      "displayName": "gemini-3.8-flash-tts",
+      "family": "gemini-3.8-flash-tts",
+      "match": {
+        "exact": [
+          "gemini-3.8-flash-tts"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "tts"
+        ],
+        "serviceType": [
+          "tts"
+        ],
+        "description": "gemini-3.8-flash-tts，官方模型目录；原生音频协议需客户端适配",
+        "contextWindow": 8192,
+        "maxOutputTokens": 16384
+      }
+    },
     {
       "id": "gemini-3.8-flash",
       "displayName": "Gemini 3.8 Flash",

compute/model-specs/kling.json

@@ -1,6 +1,46 @@
 {
   "description": "快手可灵视频生成系列模型规格。参数来源：config-center compute/providers/kling.json。所有模型均为视频生成（video_gen），无上下文窗口概念。",
   "specs": [
+    {
+      "id": "kling-3.0-turbo",
+      "displayName": "可灵 3.0 Turbo",
+      "family": "kling-3.0-turbo",
+      "match": {
+        "exact": [
+          "kling-3.0-turbo"
+        ]
+      },
+      "spec": {
+        "serviceType": [
+          "video_gen"
+        ],
+        "capabilities": [
+          "video_generation",
+          "image_to_video"
+        ],
+        "description": "可灵 3.0 Turbo，新版 /image-to-video/kling-3.0-turbo API，需独立协议适配"
+      }
+    },
+    {
+      "id": "kling-v3",
+      "displayName": "可灵 3.0",
+      "family": "kling-v3",
+      "match": {
+        "exact": [
+          "kling-v3"
+        ]
+      },
+      "spec": {
+        "serviceType": [
+          "video_gen"
+        ],
+        "capabilities": [
+          "video_generation",
+          "image_to_video"
+        ],
+        "description": "可灵 3.0，通过现有 /v1/videos/text2video 任务 API 的 model_name 调用"
+      }
+    },
     {
       "id": "kling-v2-5-turbo",
       "displayName": "可灵 V2.5 Turbo",

compute/model-specs/minimax.json

@@ -1,6 +1,86 @@
 {
   "description": "MiniMax 系列模型规格。参数来源：config-center compute/providers/minimax.json。id 保留上游 model_name 原始写法（含大小写与点号），匹配时由匹配器归一化处理。",
   "specs": [
+    {
+      "id": "MiniMax-H3-Max",
+      "displayName": "MiniMax H3 Max",
+      "family": "MiniMax-H3-Max",
+      "match": {
+        "exact": [
+          "MiniMax-H3-Max"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "image_to_video"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "MiniMax H3 Max，使用原生视频 V2 API（/v1/video_generation_v2），不适用 Anthropic Messages",
+        "extra": {
+          "maxVideoDuration": 15
+        }
+      }
+    },
+    {
+      "id": "MiniMax-H3",
+      "displayName": "MiniMax H3",
+      "family": "MiniMax-H3",
+      "match": {
+        "exact": [
+          "MiniMax-H3"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "video_generation",
+          "image_to_video"
+        ],
+        "serviceType": [
+          "video_gen"
+        ],
+        "description": "MiniMax H3，使用原生视频 V2 API（/v1/video_generation_v2），不适用 Anthropic Messages",
+        "extra": {
+          "maxVideoDuration": 15
+        }
+      }
+    },
+    {
+      "id": "MiniMax-M3.1-Flash-Preview",
+      "displayName": "MiniMax M3.1 Flash Preview",
+      "family": "MiniMax-M3.1-Flash-Preview",
+      "match": {
+        "exact": [
+          "MiniMax-M3.1-Flash-Preview"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "video_understanding",
+          "tool_use",
+          "agent",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning",
+          "vision"
+        ],
+        "description": "MiniMax M3.1 Flash Preview",
+        "contextWindow": 1000000,
+        "extra": {
+          "adaptiveThinking": true,
+          "thinkingOnly": true
+        },
+        "supportsReasoning": true
+      }
+    },
     {
       "id": "MiniMax-M3",
       "displayName": "MiniMax M3",

compute/model-specs/mistral.json

@@ -4,13 +4,12 @@
     {
       "id": "mistral-medium-3.5",
       "displayName": "Mistral Medium 3.5",
-      "family": "mistral-medium",
+      "family": "mistral-medium-3-5",
       "match": {
         "exact": [
-          "mistral-medium-3.5"
+          "mistral-medium-3-5",
-        ],
+          "mistral-medium-3.5",
-        "patterns": [
+          "mistral-medium-latest"
-          "mistral-medium-3.5*"
         ]
       },
       "spec": {
@@ -27,21 +26,21 @@
         ],
         "defaultTemperature": 1,
         "supportsReasoning": true,
-        "description": "Mistral Medium 3.5，128B 指令模型"
+        "description": "Mistral Medium 3.5，官方 API ID mistral-medium-3-5，256K 上下文"
       }
     },
     {
       "id": "mistral-large-latest",
       "displayName": "Mistral Large",
-      "family": "mistral-large",
+      "family": "mistral-large-2512",
       "match": {
-        "patterns": [
+        "exact": [
-          "mistral-large*"
+          "mistral-large-latest",
+          "mistral-large-2512"
         ]
       },
       "spec": {
         "contextWindow": 256000,
-        "maxOutputTokens": 262144,
         "capabilities": [
           "chat",
           "reasoning",
@@ -58,16 +57,16 @@
     },
     {
       "id": "mistral-small-latest",
-      "displayName": "Mistral Small",
+      "displayName": "Mistral Small 4",
-      "family": "mistral-small",
+      "family": "mistral-small-2603",
       "match": {
-        "patterns": [
+        "exact": [
-          "mistral-small*"
+          "mistral-small-latest",
+          "mistral-small-2603"
         ]
       },
       "spec": {
-        "contextWindow": 128000,
+        "contextWindow": 262144,
-        "maxOutputTokens": 8192,
         "capabilities": [
           "chat",
           "code",

compute/model-specs/moonshot.json

@@ -7,15 +7,13 @@
       "family": "kimi-k3",
       "match": {
         "exact": [
-          "kimi-k3"
+          "kimi-k3",
-        ],
+          "moonshotai/kimi-k3"
-        "patterns": [
-          "kimi-k3*"
         ]
       },
       "spec": {
         "contextWindow": 1048576,
-        "maxOutputTokens": 131072,
+        "maxOutputTokens": 1048576,
         "capabilities": [
           "chat",
           "reasoning",
@@ -23,13 +21,18 @@
           "tool_use",
           "agent",
           "long_context",
-          "vision"
+          "vision",
+          "video_understanding"
         ],
         "serviceType": [
           "reasoning"
         ],
         "supportsReasoning": true,
-        "description": "Moonshot Kimi K3 — 2.8T 参数多模态推理模型，默认 reasoning_effort=max，始终启用思考"
+        "description": "Kimi K3，1M 上下文，始终开启思考，low/high/max 三档（默认 max）；输出上限与输入共享窗口，默认 max_completion_tokens=131072",
+        "extra": {
+          "thinkingOnly": true,
+          "samplingParametersDeprecated": true
+        }
       },
       "routing": {
         "tier": "flagship",
@@ -39,14 +42,11 @@
         "reasoning": {
           "supportedModes": [
             "auto",
-            "off",
-            "minimal",
             "low",
-            "medium",
             "high",
-            "xhigh"
+            "max"
           ],
-          "defaultMode": "high"
+          "defaultMode": "max"
         }
       }
     },

compute/model-specs/openai.json

@@ -7,7 +7,8 @@
       "family": "gpt-6.1-sol",
       "match": {
         "exact": [
-          "gpt-6.1-sol"
+          "gpt-6.1-sol",
+          "openai/gpt-6.1-sol"
         ]
       },
       "spec": {

compute/model-specs/qwen.json

@@ -1,6 +1,165 @@
 {
   "description": "阿里通义千问 Qwen 系列模型规格。参数来源：config-center compute/providers/dashscope.json。",
   "specs": [
+    {
+      "id": "qwen3.8-max-prime",
+      "displayName": "Qwen3.8 Max Prime",
+      "family": "qwen3.8-max-prime",
+      "match": {
+        "exact": [
+          "qwen3.8-max-prime",
+          "qwen/qwen3.8-max-prime"
+        ]
+      },
+      "spec": {
+        "contextWindow": 1000000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "tool_use",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "supportsReasoning": true,
+        "description": "Qwen3.8 Max Prime；OpenRouter 官方目录已提供该 ID，不推断国内 DashScope 同名接入"
+      }
+    },
+    {
+      "id": "qwen-image-3.0",
+      "displayName": "qwen-image-3.0",
+      "family": "qwen-image-3.0",
+      "match": {
+        "exact": [
+          "qwen-image-3.0"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "image_generation",
+          "image_editing",
+          "text_rendering"
+        ],
+        "serviceType": [
+          "image_gen"
+        ],
+        "description": "qwen-image-3.0"
+      }
+    },
+    {
+      "id": "qwen-image-3.0-pro",
+      "displayName": "qwen-image-3.0-pro",
+      "family": "qwen-image-3.0-pro",
+      "match": {
+        "exact": [
+          "qwen-image-3.0-pro"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "image_generation",
+          "image_editing",
+          "text_rendering"
+        ],
+        "serviceType": [
+          "image_gen"
+        ],
+        "description": "qwen-image-3.0-pro"
+      }
+    },
+    {
+      "id": "qwen3.7-text-rerank",
+      "displayName": "Qwen3.7 Text Rerank",
+      "family": "qwen3.7-text-rerank",
+      "match": {
+        "exact": [
+          "qwen3.7-text-rerank"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "rerank",
+          "multilingual"
+        ],
+        "serviceType": [
+          "rerank"
+        ],
+        "description": "Qwen3.7 Text Rerank"
+      }
+    },
+    {
+      "id": "qwen3.7-text-embedding-flash",
+      "displayName": "qwen3.7-text-embedding-flash",
+      "family": "qwen3.7-text-embedding-flash",
+      "match": {
+        "exact": [
+          "qwen3.7-text-embedding-flash"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "text_embedding",
+          "multilingual"
+        ],
+        "serviceType": [
+          "embedding"
+        ],
+        "description": "qwen3.7-text-embedding-flash"
+      }
+    },
+    {
+      "id": "qwen3.7-text-embedding",
+      "displayName": "qwen3.7-text-embedding",
+      "family": "qwen3.7-text-embedding",
+      "match": {
+        "exact": [
+          "qwen3.7-text-embedding"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "text_embedding",
+          "multilingual"
+        ],
+        "serviceType": [
+          "embedding"
+        ],
+        "description": "qwen3.7-text-embedding"
+      }
+    },
+    {
+      "id": "qwen3.8-omni-flash",
+      "displayName": "Qwen3.8 Omni Flash",
+      "family": "qwen3.8-omni-flash",
+      "match": {
+        "exact": [
+          "qwen3.8-omni-flash",
+          "qwen/qwen3.8-omni-flash"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "vision",
+          "audio_understanding",
+          "video_understanding",
+          "tool_use",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat",
+          "vision"
+        ],
+        "description": "Qwen3.8 Omni Flash",
+        "contextWindow": 1000000,
+        "maxOutputTokens": 131072,
+        "supportsReasoning": true
+      }
+    },
     {
       "id": "qwen3.8-max",
       "displayName": "Qwen3.8 Max",

compute/model-specs/stability.json

@@ -1,6 +1,27 @@
 {
   "description": "Stability AI 图像生成系列模型规格。参数来源：config-center compute/providers/stability.json。无上下文窗口（图像生成模型）。",
   "specs": [
+    {
+      "id": "stable-audio-3.0",
+      "displayName": "Stable Audio 3.0",
+      "family": "stable-audio-3.0",
+      "match": {
+        "exact": [
+          "stable-audio-3.0"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "music_generation",
+          "audio_generation"
+        ],
+        "serviceType": [
+          "music_gen"
+        ],
+        "releasedAt": "2026-05-20",
+        "description": "Stable Audio 3.0，通过原生 /v2beta/audio/stable-audio API 提供，不使用 Stability 图像协议"
+      }
+    },
     {
       "id": "stable-diffusion-3.5-large",
       "displayName": "Stable Diffusion 3.5 Large",

compute/model-specs/tencent.json

@@ -1,6 +1,33 @@
 {
   "description": "腾讯混元系列模型规格。参数来源：config-center compute/providers/tencent.json。hunyuan-2.0-thinking 和 hunyuan-t1 为推理模型。",
   "specs": [
+    {
+      "id": "hy4-preview",
+      "displayName": "Tencent Hy4 Preview",
+      "family": "hy4-preview",
+      "match": {
+        "exact": [
+          "hy4-preview",
+          "tencent/hy4-preview"
+        ]
+      },
+      "spec": {
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "tool_use",
+          "agent",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat"
+        ],
+        "description": "Tencent Hy4 Preview，通过 TokenHub 提供；既有 Hunyuan Provider 端点可用性尚未确认",
+        "contextWindow": 1048576,
+        "releasedAt": "2026-08-28"
+      }
+    },
     {
       "id": "hy3",
       "displayName": "混元 Hy3",

compute/model-specs/volcengine.json

@@ -1,6 +1,66 @@
 {
   "description": "火山引擎豆包系列模型规格。参数来源：config-center compute/providers/volcengine.json。包含豆包 Seed 系列对话、DeepSeek 托管版、Embedding、语音克隆、实时语音、同传和翻译服务。doubao-seed-1.6-thinking 为推理模型。",
   "specs": [
+    {
+      "id": "doubao-seed-evolving",
+      "displayName": "Doubao Seed Evolving",
+      "family": "doubao-seed-evolving",
+      "match": {
+        "exact": [
+          "doubao-seed-evolving"
+        ]
+      },
+      "spec": {
+        "contextWindow": 1000000,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "tool_use",
+          "agent",
+          "long_context"
+        ],
+        "serviceType": [
+          "chat",
+          "reasoning"
+        ],
+        "supportsReasoning": true,
+        "description": "Seed 系列持续迭代模型，通过固定 ID 获取最新版本，2026-09-09 更新"
+      }
+    },
+    {
+      "id": "doubao-seed-2.1-pro",
+      "displayName": "豆包 Seed 2.1 Pro",
+      "family": "doubao-seed-2.1-pro",
+      "match": {
+        "exact": [
+          "doubao-seed-2.1-pro"
+        ]
+      },
+      "spec": {
+        "serviceType": [
+          "chat",
+          "reasoning",
+          "vision"
+        ],
+        "description": "火山方舟豆包 Seed 2.1 Pro，生产级 Coding、Agent 与多模态旗舰模型",
+        "contextWindow": 256000,
+        "maxOutputTokens": 262144,
+        "capabilities": [
+          "chat",
+          "reasoning",
+          "code",
+          "vision",
+          "multilingual",
+          "long_context",
+          "tool_use",
+          "agent",
+          "deep_thinking"
+        ],
+        "supportsReasoning": true
+      }
+    },
     {
       "id": "doubao-seed-2.0-pro",
       "displayName": "豆包 2.0 Pro",

compute/model-specs/zhipu.json

@@ -101,7 +101,7 @@
             "high",
             "max"
           ],
-          "defaultMode": "high"
+          "defaultMode": "max"
         }
       }
     },
@@ -157,7 +157,7 @@
             "high",
             "max"
           ],
-          "defaultMode": "high"
+          "defaultMode": "max"
         }
       }
     },

compute/providers/cohere.json

@@ -15,20 +15,95 @@
     "rerank"
   ],
   "models": [
+    {
+      "modelName": "embed-v5.0-fast",
+      "displayName": "Embed 5 Fast",
+      "capabilities": [
+        "text_embedding",
+        "multilingual"
+      ],
+      "serviceType": [
+        "embedding"
+      ],
+      "description": "Embed 5 Fast；当前 OpenAI 兼容接入仅声明文本向量能力",
+      "contextWindow": 128000,
+      "extra": {
+        "dimensions": [
+          256,
+          512,
+          768,
+          1024,
+          1536,
+          2048
+        ],
+        "defaultDimension": 2048,
+        "pricingNotes": "Token 单价未由当前官方价格页公开确认；Model Vault 按实例计价。"
+      }
+    },
+    {
+      "modelName": "embed-v5.0-pro",
+      "displayName": "Embed 5 Pro",
+      "capabilities": [
+        "text_embedding",
+        "multilingual"
+      ],
+      "serviceType": [
+        "embedding"
+      ],
+      "description": "Embed 5 Pro；当前 OpenAI 兼容接入仅声明文本向量能力",
+      "contextWindow": 128000,
+      "extra": {
+        "dimensions": [
+          256,
+          512,
+          768,
+          1024,
+          1536,
+          2048
+        ],
+        "defaultDimension": 2048,
+        "pricingNotes": "Token 单价未由当前官方价格页公开确认；Model Vault 按实例计价。"
+      }
+    },
+    {
+      "modelName": "command-a-plus-05-2026",
+      "displayName": "Command A+",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "vision",
+        "tool_use",
+        "multilingual",
+        "long_context"
+      ],
+      "serviceType": [
+        "chat"
+      ],
+      "description": "Command A+",
+      "contextWindow": 128000,
+      "maxOutputTokens": 64000,
+      "extra": {
+        "pricingNotes": "官方文档声明限流额度内免费；生产 Model Vault 按实例单独计价，不预填 token 单价。",
+        "reasoning": {
+          "supportedEfforts": [
+            "none",
+            "high"
+          ]
+        }
+      }
+    },
     {
       "modelName": "command-a-03-2025",
       "displayName": "Command A",
       "serviceType": [
         "chat"
       ],
-      "description": "Cohere 旗舰对话模型，256K 上下文，擅长 RAG 和工具调用",
+      "description": "Cohere Command A 文本模型，256K 上下文，支持 RAG 和工具调用",
       "contextWindow": 256000,
       "maxOutputTokens": 8000,
       "capabilities": [
         "chat",
-        "reasoning",
         "code",
-        "vision",
         "tool_use",
         "rag",
         "long_context"

compute/providers/dashscope.json

@@ -34,6 +34,103 @@
     "qwen3-max-trans"
   ],
   "models": [
+    {
+      "modelName": "qwen-image-3.0",
+      "displayName": "qwen-image-3.0",
+      "capabilities": [
+        "image_generation",
+        "image_editing",
+        "text_rendering"
+      ],
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "qwen-image-3.0",
+      "extra": {
+        "pricingNotes": "按图片计费，北京：输入 0.02 元/张；Pro 输出 1K 0.25 元、2K 0.5 元；普通版输出 0.18 元/张。"
+      }
+    },
+    {
+      "modelName": "qwen-image-3.0-pro",
+      "displayName": "qwen-image-3.0-pro",
+      "capabilities": [
+        "image_generation",
+        "image_editing",
+        "text_rendering"
+      ],
+      "serviceType": [
+        "image_gen"
+      ],
+      "description": "qwen-image-3.0-pro",
+      "extra": {
+        "pricingNotes": "按图片计费，北京：输入 0.02 元/张；Pro 输出 1K 0.25 元、2K 0.5 元；普通版输出 0.18 元/张。"
+      }
+    },
+    {
+      "modelName": "qwen3.7-text-rerank",
+      "displayName": "Qwen3.7 Text Rerank",
+      "capabilities": [
+        "rerank",
+        "multilingual"
+      ],
+      "serviceType": [
+        "rerank"
+      ],
+      "description": "Qwen3.7 Text Rerank"
+    },
+    {
+      "modelName": "qwen3.7-text-embedding-flash",
+      "displayName": "qwen3.7-text-embedding-flash",
+      "capabilities": [
+        "text_embedding",
+        "multilingual"
+      ],
+      "serviceType": [
+        "embedding"
+      ],
+      "description": "qwen3.7-text-embedding-flash",
+      "inputPrice": 0.125,
+      "outputPrice": 0
+    },
+    {
+      "modelName": "qwen3.7-text-embedding",
+      "displayName": "qwen3.7-text-embedding",
+      "capabilities": [
+        "text_embedding",
+        "multilingual"
+      ],
+      "serviceType": [
+        "embedding"
+      ],
+      "description": "qwen3.7-text-embedding",
+      "inputPrice": 0.5,
+      "outputPrice": 0
+    },
+    {
+      "modelName": "qwen3.8-omni-flash",
+      "displayName": "Qwen3.8 Omni Flash",
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "vision",
+        "audio_understanding",
+        "video_understanding",
+        "tool_use",
+        "long_context"
+      ],
+      "serviceType": [
+        "chat",
+        "vision"
+      ],
+      "description": "Qwen3.8 Omni Flash",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 131072,
+      "inputPrice": 0.8,
+      "outputPrice": 2.7,
+      "extra": {
+        "cachedInputPrice": 0.1
+      }
+    },
     {
       "modelName": "qwen3.8-max",
       "displayName": "阿里云 Qwen3.8-Max",

compute/providers/deepseek.json

@@ -51,6 +51,7 @@
         "thinkingDefault": true,
         "reasoning": {
           "supportedEfforts": [
+            "low",
             "high",
             "max"
           ],
@@ -109,15 +110,19 @@
         "concurrencyLimit": 2500,
         "supportsThinking": true,
         "thinkingDefault": true,
-        "reasoningEffort": [
-          "high",
-          "max"
-        ],
         "features": {
           "jsonOutput": true,
           "toolCalls": true,
           "chatPrefixCompletion": "beta",
           "fimCompletion": false
+        },
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "high"
         }
       }
     },
@@ -149,15 +154,19 @@
         "concurrencyLimit": 2500,
         "supportsThinking": true,
         "thinkingDefault": true,
-        "reasoningEffort": [
-          "high",
-          "max"
-        ],
         "features": {
           "jsonOutput": true,
           "toolCalls": true,
           "chatPrefixCompletion": "beta",
           "fimCompletion": "non-thinking-only"
+        },
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "high"
         }
       }
     },
@@ -190,15 +199,19 @@
         "concurrencyLimit": 500,
         "supportsThinking": true,
         "thinkingDefault": true,
-        "reasoningEffort": [
-          "high",
-          "max"
-        ],
         "features": {
           "jsonOutput": true,
           "toolCalls": true,
           "chatPrefixCompletion": "beta",
           "fimCompletion": "non-thinking-only"
+        },
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "high"
         }
       }
     }

compute/providers/google.json

@@ -15,6 +15,30 @@
     "embedding"
   ],
   "models": [
+    {
+      "modelName": "gemini-embedding-2",
+      "displayName": "Gemini Embedding 2",
+      "contextWindow": 8192,
+      "capabilities": [
+        "text_embedding",
+        "vision",
+        "audio_understanding",
+        "video_understanding",
+        "multilingual"
+      ],
+      "serviceType": [
+        "embedding"
+      ],
+      "description": "Gemini Embedding 2，原生多模态嵌入，8192 token 输入，共享文本/图像/音视频向量空间",
+      "extra": {
+        "dimensions": [
+          768,
+          1536,
+          3072
+        ],
+        "defaultDimension": 3072
+      }
+    },
     {
       "modelName": "gemini-3.8-flash",
       "displayName": "Gemini 3.8 Flash",

compute/providers/kling.json

@@ -13,6 +13,18 @@
     "video_gen"
   ],
   "models": [
+    {
+      "modelName": "kling-v3",
+      "displayName": "可灵 3.0",
+      "serviceType": [
+        "video_gen"
+      ],
+      "capabilities": [
+        "video_generation",
+        "image_to_video"
+      ],
+      "description": "可灵 3.0，官方 legacy /v1/videos/text2video 协议已支持该 model_name"
+    },
     {
       "modelName": "kling-v2-5-turbo",
       "displayName": "可灵 V2.5 Turbo",

compute/providers/mistral.json

@@ -14,6 +14,25 @@
     "fast"
   ],
   "models": [
+    {
+      "modelName": "mistral-medium-3-5",
+      "displayName": "Mistral Medium 3.5",
+      "contextWindow": 262144,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use"
+      ],
+      "serviceType": [
+        "chat"
+      ],
+      "defaultTemperature": 1,
+      "description": "Mistral Medium 3.5，官方 API ID mistral-medium-3-5，256K 上下文",
+      "inputPrice": 1.5,
+      "outputPrice": 7.5
+    },
     {
       "modelName": "mistral-medium-latest",
       "displayName": "Mistral Medium 3.5",
@@ -21,8 +40,7 @@
         "chat"
       ],
       "description": "Mistral 前沿多模态模型，面向 Agentic 和 Coding 场景优化，256K 上下文",
-      "contextWindow": 256000,
+      "contextWindow": 262144,
-      "maxOutputTokens": 262144,
       "capabilities": [
         "chat",
         "reasoning",
@@ -46,7 +64,6 @@
       ],
       "description": "Mistral 旗舰模型，256K 上下文，支持视觉和工具调用",
       "contextWindow": 256000,
-      "maxOutputTokens": 262144,
       "capabilities": [
         "chat",
         "reasoning",
@@ -67,9 +84,8 @@
       "serviceType": [
         "fast"
       ],
-      "description": "Mistral Small 4，统一指令、推理和编码能力的高效多模态模型，256K 上下文",
+      "description": "Mistral Small 4，指令、推理与编码混合模型，256K 上下文",
-      "contextWindow": 256000,
+      "contextWindow": 262144,
-      "maxOutputTokens": 8192,
       "capabilities": [
         "chat",
         "code",

compute/providers/moonshot.json

@@ -25,6 +25,44 @@
     "kimi-thinking-preview"
   ],
   "models": [
+    {
+      "modelName": "kimi-k3",
+      "displayName": "Kimi K3",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 1048576,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "tool_use",
+        "agent",
+        "long_context",
+        "vision",
+        "video_understanding"
+      ],
+      "serviceType": [
+        "reasoning"
+      ],
+      "description": "Kimi K3，1M 上下文，始终开启思考，low/high/max 三档（默认 max）；输出上限与输入共享窗口，默认 max_completion_tokens=131072",
+      "extra": {
+        "thinkingOnly": true,
+        "thinking": {
+          "default": "enabled",
+          "disableSupported": false,
+          "preserveReasoningContent": true
+        },
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "max"
+        },
+        "pricingNotes": "国内官方价格表当前未返回模型单价；不将国际美元价格换算为国内人民币价格。",
+        "samplingParametersDeprecated": true
+      }
+    },
     {
       "modelName": "kimi-k2.7-code",
       "displayName": "Kimi K2.7 Code",

compute/providers/openrouter.json

@@ -15,6 +15,280 @@
     "vision"
   ],
   "models": [
+    {
+      "modelName": "cohere/command-a-plus",
+      "displayName": "Cohere: Command A+",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 192000,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 0.3,
+      "outputPrice": 1.5,
+      "description": "OpenRouter 官方公共目录模型，接入参数与计价以该平台为准",
+      "maxOutputTokens": 64000
+    },
+    {
+      "modelName": "mistralai/mistral-medium-3-5",
+      "displayName": "Mistral: Mistral Medium 3.5",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 262144,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 1.5,
+      "outputPrice": 7.5,
+      "description": "OpenRouter 官方公共目录模型，接入参数与计价以该平台为准",
+      "maxOutputTokens": 209715
+    },
+    {
+      "modelName": "google/gemini-3.8-flash",
+      "displayName": "Google: Gemini 3.8 Flash",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1048576,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 0.75,
+      "outputPrice": 3.75,
+      "description": "OpenRouter 官方公共目录模型，接入参数与计价以该平台为准",
+      "maxOutputTokens": 65536
+    },
+    {
+      "modelName": "minimax/minimax-m3",
+      "displayName": "MiniMax: MiniMax M3",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1048576,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 0.3,
+      "outputPrice": 1.2,
+      "description": "OpenRouter 官方公共目录模型，接入参数与计价以该平台为准",
+      "maxOutputTokens": 512000
+    },
+    {
+      "modelName": "qwen/qwen3.8-omni-flash",
+      "displayName": "Qwen: Qwen3.8 Omni Flash",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1000000,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 0.15,
+      "outputPrice": 0.47,
+      "description": "OpenRouter 官方公共目录模型，接入参数与计价以该平台为准",
+      "maxOutputTokens": 131072
+    },
+    {
+      "modelName": "qwen/qwen3.8-max-prime",
+      "displayName": "Qwen: Qwen3.8 Max Prime",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1000000,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 4.0,
+      "outputPrice": 12.0,
+      "description": "OpenRouter 官方公共目录模型，接入参数与计价以该平台为准",
+      "maxOutputTokens": 131072
+    },
+    {
+      "modelName": "deepseek/deepseek-v4.1-flash",
+      "displayName": "DeepSeek: DeepSeek V4.1 Flash",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1048576,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 0.3,
+      "outputPrice": 1.2,
+      "description": "OpenRouter 官方公共目录模型；接入参数与计价以该平台为准",
+      "extra": {
+        "pricingNotes": "USD per 1M tokens; OpenRouter access-specific pricing."
+      },
+      "maxOutputTokens": 943718
+    },
+    {
+      "modelName": "tencent/hy4-preview",
+      "displayName": "Tencent: Hy4 preview",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1048576,
+      "capabilities": [
+        "chat",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 0.834,
+      "outputPrice": 2.501,
+      "description": "OpenRouter 官方公共目录模型；接入参数与计价以该平台为准",
+      "extra": {
+        "pricingNotes": "USD per 1M tokens; OpenRouter access-specific pricing."
+      },
+      "maxOutputTokens": 64000
+    },
+    {
+      "modelName": "z-ai/glm-5.3-flash",
+      "displayName": "Z.ai: GLM 5.3 Flash",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1048576,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 0.15,
+      "outputPrice": 0.5,
+      "description": "OpenRouter 官方公共目录模型；接入参数与计价以该平台为准",
+      "extra": {
+        "pricingNotes": "USD per 1M tokens; OpenRouter access-specific pricing."
+      },
+      "maxOutputTokens": 943717
+    },
+    {
+      "modelName": "z-ai/glm-5.3",
+      "displayName": "Z.ai: GLM 5.3",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1048576,
+      "capabilities": [
+        "chat",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 1.4,
+      "outputPrice": 4.4,
+      "description": "OpenRouter 官方公共目录模型；接入参数与计价以该平台为准",
+      "extra": {
+        "pricingNotes": "USD per 1M tokens; OpenRouter access-specific pricing."
+      },
+      "maxOutputTokens": 131072
+    },
+    {
+      "modelName": "moonshotai/kimi-k3",
+      "displayName": "MoonshotAI: Kimi K3",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1048576,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 2.7,
+      "outputPrice": 13.5,
+      "description": "OpenRouter 官方公共目录模型；接入参数与计价以该平台为准",
+      "extra": {
+        "pricingNotes": "USD per 1M tokens; OpenRouter access-specific pricing."
+      },
+      "maxOutputTokens": 943718
+    },
+    {
+      "modelName": "anthropic/claude-sonnet-5.5",
+      "displayName": "Anthropic: Claude Sonnet 5.5",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1000000,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 2.0,
+      "outputPrice": 10.0,
+      "description": "OpenRouter 官方公共目录模型；接入参数与计价以该平台为准",
+      "extra": {
+        "pricingNotes": "USD per 1M tokens; OpenRouter access-specific pricing."
+      },
+      "maxOutputTokens": 128000
+    },
+    {
+      "modelName": "anthropic/claude-opus-5.5",
+      "displayName": "Anthropic: Claude Opus 5.5",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1000000,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 4.0,
+      "outputPrice": 20.0,
+      "description": "OpenRouter 官方公共目录模型；接入参数与计价以该平台为准",
+      "extra": {
+        "pricingNotes": "USD per 1M tokens; OpenRouter access-specific pricing."
+      },
+      "maxOutputTokens": 128000
+    },
+    {
+      "modelName": "openai/gpt-6.1-sol",
+      "displayName": "OpenAI: GPT-6.1 Sol",
+      "serviceType": [
+        "chat"
+      ],
+      "contextWindow": 1050000,
+      "capabilities": [
+        "chat",
+        "vision",
+        "tool_use",
+        "reasoning"
+      ],
+      "inputPrice": 2.0,
+      "outputPrice": 10.0,
+      "description": "OpenRouter 官方公共目录模型；接入参数与计价以该平台为准",
+      "extra": {
+        "pricingNotes": "USD per 1M tokens; OpenRouter access-specific pricing."
+      },
+      "maxOutputTokens": 128000
+    },
     {
       "modelName": "stealth/ox-alpha",
       "displayName": "Ox Alpha (免费)",

compute/providers/volcengine.json

@@ -22,6 +22,25 @@
     "video_gen"
   ],
   "models": [
+    {
+      "modelName": "doubao-seed-evolving",
+      "displayName": "Doubao Seed Evolving",
+      "contextWindow": 1000000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "vision",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "serviceType": [
+        "chat",
+        "reasoning"
+      ],
+      "description": "Seed 系列持续迭代模型，通过固定 ID 获取最新版本，2026-09-09 更新"
+    },
     {
       "modelName": "doubao-seed-2.1-pro",
       "displayName": "豆包 Seed 2.1 Pro",

compute/providers/zhipu.json

@@ -25,6 +25,125 @@
     "glm-4-0520"
   ],
   "models": [
+    {
+      "modelName": "glm-5.3-flashx",
+      "displayName": "GLM-5.3-FlashX",
+      "contextWindow": 1000000,
+      "maxOutputTokens": 128000,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "code",
+        "multilingual",
+        "vision",
+        "video_understanding",
+        "tool_use",
+        "long_context",
+        "fast"
+      ],
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "description": "Z.ai GLM-5.3-FlashX，高速多模态推理模型，100 万上下文，始终开启推理",
+      "extra": {
+        "thinkingOnly": true,
+        "cacheHitPrice": 0.57,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "max"
+        },
+        "pricingNotes": "国内官方标准价，人民币/百万 tokens；Flash 限时优惠以控制台实际结算为准。"
+      },
+      "inputPrice": 2,
+      "outputPrice": 7
+    },
+    {
+      "modelName": "glm-5.3-flash",
+      "displayName": "GLM-5.3-Flash",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "tool_use",
+        "agent",
+        "long_context",
+        "vision",
+        "image_understanding",
+        "video_understanding"
+      ],
+      "serviceType": [
+        "chat",
+        "reasoning",
+        "vision"
+      ],
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "description": "智谱 GLM-5.3-Flash 原生多模态模型，混合稀疏与线性注意力，面向高效编码与长程 Agent 任务，100 万上下文；强制思考，无法关闭",
+      "extra": {
+        "thinkingDefault": true,
+        "thinkingOnly": true,
+        "cacheHitPrice": 0.23,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "max"
+        },
+        "pricingNotes": "国内官方标准价，人民币/百万 tokens；Flash 限时优惠以控制台实际结算为准。"
+      },
+      "inputPrice": 0.8,
+      "outputPrice": 2.8
+    },
+    {
+      "modelName": "glm-5.3",
+      "displayName": "GLM-5.3",
+      "contextWindow": 1048576,
+      "maxOutputTokens": 131072,
+      "capabilities": [
+        "chat",
+        "reasoning",
+        "deep_thinking",
+        "code",
+        "math",
+        "multilingual",
+        "tool_use",
+        "agent",
+        "long_context"
+      ],
+      "serviceType": [
+        "chat"
+      ],
+      "defaultTemperature": 1,
+      "defaultTopP": 0.95,
+      "description": "智谱 GLM-5.3 大推理模型，面向复杂软件工程与长程 Agent 任务，100 万上下文；强制思考，无法关闭",
+      "extra": {
+        "thinkingDefault": true,
+        "thinkingOnly": true,
+        "cacheHitPrice": 2,
+        "reasoning": {
+          "supportedEfforts": [
+            "low",
+            "high",
+            "max"
+          ],
+          "defaultEffort": "max"
+        },
+        "pricingNotes": "国内官方标准价，人民币/百万 tokens；Flash 限时优惠以控制台实际结算为准。"
+      },
+      "inputPrice": 8,
+      "outputPrice": 28
+    },
     {
       "modelName": "glm-5.2",
       "displayName": "GLM-5.2",

manifest.json

@@ -1,6 +1,6 @@
 {
   "version": "1.0.0",
-  "presetDataVersion": 123,
+  "presetDataVersion": 124,
-  "updatedAt": "2026-10-01",
+  "updatedAt": "2026-10-02",
   "description": "DesireCore 官方配置中心"
 }

reports/provider-model-audit-2026-10-02.json

@@ -0,0 +1,461 @@
+{
+  "checkedAt": "2026-10-02",
+  "cutoff": "2025-10-02",
+  "scope": {
+    "providers": 28,
+    "codingPlans": 11
+  },
+  "method": "逐项检索官方模型目录/发布记录；模型发布日期与本仓文件更新时间分别统计。完整最新名单、接入端点或单价无法确认时明确保留疑点。配置和 schema 校验不等同真实账号调用验收。",
+  "confirmedSuppliersWithoutModelUpdatesForOneYear": [],
+  "inactivityConclusion": "未找到足够官方证据确认某个自研供应商在整个过去一年没有任何模型更新。来源受限或无完整日期记录的项目不能按无更新判定。",
+  "stalePresetFamilies": [
+    {
+      "config": "compute/providers/stability.json",
+      "family": "Stable Diffusion 3.5",
+      "releasedAt": "2024-10-22",
+      "source": "https://stability.ai/news-updates/introducing-stable-diffusion-3-5",
+      "note": "图像代际超过一年；Stability 在 2026-05-20 发布 Stable Audio 3.0，供应商整体仍更新。"
+    },
+    {
+      "config": "compute/providers/local-whisper.json",
+      "family": "Whisper large-v3",
+      "releasedAt": "2023-11-07",
+      "source": "https://huggingface.co/api/models/openai/whisper-large-v3",
+      "note": "这是本地模型预置，不是独立供应商；后继 large-v3-turbo 2024-10-01 已发布，未假定本地服务已安装。"
+    },
+    {
+      "config": "compute/providers/mistral.json",
+      "family": "Codestral 25.08",
+      "releasedAt": "2025-07-30",
+      "source": "https://docs.mistral.ai/models/codestral-25-08",
+      "note": "编程补全代际超过一年；Mistral Medium 3.5 和 Small 4 在 2026 已发布。"
+    }
+  ],
+  "entries": [
+    {
+      "config": "compute/providers/openai.json",
+      "name": "openai",
+      "source": "https://developers.openai.com/api/docs/models",
+      "finding": "GPT-6.1 Sol / GPT Image 2.5，已在上次发布补齐",
+      "newPresetIds": [],
+      "modelCount": 37,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/providers/openai-codex.json",
+      "name": "openai-codex",
+      "source": "https://developers.openai.com/api/docs/models/gpt-6.1-sol",
+      "finding": "GPT-6.1 Sol 已存在，保留订阅推理档位",
+      "newPresetIds": [],
+      "modelCount": 12,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/providers/anthropic.json",
+      "name": "anthropic",
+      "source": "https://platform.claude.com/docs/en/models/overview",
+      "finding": "Opus 5.5 / Sonnet 5.5 / Fable 5.1 已存在",
+      "newPresetIds": [],
+      "modelCount": 5,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/providers/anthropic-claude.json",
+      "name": "anthropic-claude",
+      "source": "https://platform.claude.com/docs/en/models/overview",
+      "finding": "订阅 Provider 已包含当前代际",
+      "newPresetIds": [],
+      "modelCount": 5,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/providers/github-copilot.json",
+      "name": "github-copilot",
+      "source": "https://docs.github.com/en/copilot/reference/ai-models/supported-models",
+      "finding": "目录由已授权账号动态发现；保持空的静态模型列表",
+      "newPresetIds": [],
+      "modelCount": 0,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/providers/deepseek.json",
+      "name": "deepseek",
+      "source": "https://api-docs.deepseek.com/guides/thinking_mode/",
+      "finding": "V4.1 Flash 已存在，新增 low 推理档并清除旧的无效 flat effort 字段",
+      "newPresetIds": [],
+      "modelCount": 4,
+      "configurationChanged": true
+    },
+    {
+      "config": "compute/providers/dashscope.json",
+      "name": "dashscope",
+      "source": "https://help.aliyun.com/zh/model-studio/models",
+      "finding": "新增 Omni Flash、Image 3.0 与向量/重排序代际",
+      "newPresetIds": [
+        "qwen-image-3.0",
+        "qwen-image-3.0-pro",
+        "qwen3.7-text-embedding",
+        "qwen3.7-text-embedding-flash",
+        "qwen3.7-text-rerank",
+        "qwen3.8-omni-flash"
+      ],
+      "modelCount": 35,
+      "configurationChanged": true
+    },
+    {
+      "config": "compute/providers/volcengine.json",
+      "name": "volcengine",
+      "source": "https://docs.volcengine.com/docs/ark/latest-model?lang=zh",
+      "finding": "Seed Evolving、Seed 2.1；补齐 Provider/shared-spec，其他套餐候选待当前端点核对",
+      "newPresetIds": [
+        "doubao-seed-evolving"
+      ],
+      "modelCount": 24,
+      "configurationChanged": true
+    },
+    {
+      "config": "compute/providers/moonshot.json",
+      "name": "moonshot",
+      "source": "https://platform.kimi.com/docs/guide/kimi-k3-quickstart",
+      "finding": "新增国内 K3 接入，纠正始终思考与默认 max；当前国内官方价格表未返回单价",
+      "newPresetIds": [
+        "kimi-k3"
+      ],
+      "modelCount": 11,
+      "configurationChanged": true
+    },
+    {
+      "config": "compute/providers/zhipu.json",
+      "name": "zhipu",
+      "source": "https://docs.bigmodel.cn/cn/guide/models/vlm/glm-5.3-flash",
+      "finding": "新增 GLM-5.3 / Flash / FlashX 的国内 API 接入，标准人民币价格已核对",
+      "newPresetIds": [
+        "glm-5.3",
+        "glm-5.3-flash",
+        "glm-5.3-flashx"
+      ],
+      "modelCount": 12,
+      "configurationChanged": true
+    },
+    {
+      "config": "compute/providers/zhipu-embedding.json",
+      "name": "zhipu-embedding",
+      "source": "https://docs.bigmodel.cn/cn/guide/start/pricing.md",
+      "finding": "未发现 embedding-3 的已验证后继 ID，保留",
+      "newPresetIds": [],
+      "modelCount": 1,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/providers/baichuan.json",
+      "name": "baichuan",
+      "source": "https://www.baichuan-ai.com/blog/baichuan-M3",
+      "finding": "M3 系列已存在；官网正文读取受限，不能保证无更新型号",
+      "newPresetIds": [],
+      "modelCount": 4,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/providers/minimax.json",
+      "name": "minimax",
+      "source": "https://platform.minimax.cn/docs/api-reference/text-anthropic-api",
+      "finding": "M3.1 Preview 仅供套餐；H3/H3 Max 新视频协议只加入共享规格",
+      "newPresetIds": [],
+      "modelCount": 26,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/providers/xiaomi.json",
+      "name": "xiaomi",
+      "source": "https://mimo.mi.com/docs/en-US/news/latest/v2-6",
+      "finding": "MiMo V2.6 系列已有，未发现已验证后继 API 代际",
+      "newPresetIds": [],
+      "modelCount": 10,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/providers/baidu.json",
+      "name": "baidu",
+      "source": "https://intl.cloud.baidu.com/en/doc/qianfan/s/7m95lyy43-intl-en",
+      "finding": "ERNIE 5.0 已有；区域文档与国内控制台存在差异，未增加猜测的 ID",
+      "newPresetIds": [],
+      "modelCount": 5,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/providers/tencent.json",
+      "name": "tencent",
+      "source": "https://www.tencent.com/tencent-releases-and-open-sources-tencent-hy4-preview/",
+      "finding": "新增 Hy4 Preview 共享规格及 OpenRouter 接入；既有 Hunyuan 端点支持尚未确认",
+      "newPresetIds": [],
+      "modelCount": 6,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/providers/xunfei.json",
+      "name": "xunfei",
+      "source": "https://xinghuo.xfyun.cn/sparkapi",
+      "finding": "官网展示 Spark X2.5；现有 spark-x 的 X2 文档要求 /x2 端点，当前 /v1 还承载 Ultra，不能直接替换；X2.5 API ID 未验证",
+      "newPresetIds": [],
+      "modelCount": 2,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/providers/siliconflow.json",
+      "name": "siliconflow",
+      "source": "https://www.siliconflow.cn/models",
+      "finding": "最新目录已检索；国际站 GLM-5.3 有发布，国内模型 ID/人民币价格待模型列表确认，未混用国际美元价",
+      "newPresetIds": [],
+      "modelCount": 3,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/providers/local-whisper.json",
+      "name": "local-whisper",
+      "source": "https://github.com/openai/whisper/releases",
+      "finding": "本地运行器；预置 large-v3 是旧模型，软件更新日期不能视作模型更新日期",
+      "newPresetIds": [],
+      "modelCount": 1,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/providers/ollama.json",
+      "name": "ollama",
+      "source": "https://ollama.com/blog",
+      "finding": "本地运行器，实际模型由 /api/tags 动态发现；保留本地安装样例，不假定新模型已安装",
+      "newPresetIds": [],
+      "modelCount": 1,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/providers/google.json",
+      "name": "google",
+      "source": "https://ai.google.dev/gemini-api/docs/models",
+      "finding": "Gemini 3.8 Flash 已有；新增 Gemini Embedding 2，音频代际补入共享规格；新 wire schema 需客户端适配",
+      "newPresetIds": [
+        "gemini-embedding-2"
+      ],
+      "modelCount": 11,
+      "configurationChanged": true
+    },
+    {
+      "config": "compute/providers/mistral.json",
+      "name": "mistral",
+      "source": "https://docs.mistral.ai/models",
+      "finding": "修正最新 API 拼写、Small 4 窗口及 Large 3 / Small 4 价格",
+      "newPresetIds": [
+        "mistral-medium-3-5"
+      ],
+      "modelCount": 5,
+      "configurationChanged": true
+    },
+    {
+      "config": "compute/providers/xai.json",
+      "name": "xai",
+      "source": "https://docs.x.ai/developers/release-notes",
+      "finding": "Grok 4.7 已存在，未发现已验证更新的通用模型代际",
+      "newPresetIds": [],
+      "modelCount": 4,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/providers/cohere.json",
+      "name": "cohere",
+      "source": "https://docs.cohere.com/docs/models",
+      "finding": "新增 Command A+ 与 Embed 5 Provider；Rerank 4 仅共享规格，需原生协议适配",
+      "newPresetIds": [
+        "command-a-plus-05-2026",
+        "embed-v5.0-fast",
+        "embed-v5.0-pro"
+      ],
+      "modelCount": 7,
+      "configurationChanged": true
+    },
+    {
+      "config": "compute/providers/kling.json",
+      "name": "kling",
+      "source": "https://kling.ai/document-api/api/video/3-0-omni/text-to-video/legacy",
+      "finding": "旧任务协议明确支持 kling-v3，已加入 Provider；3.0 Turbo 新协议仅共享规格",
+      "newPresetIds": [
+        "kling-v3"
+      ],
+      "modelCount": 5,
+      "configurationChanged": true
+    },
+    {
+      "config": "compute/providers/stability.json",
+      "name": "stability",
+      "source": "https://stability.ai/news-updates",
+      "finding": "2026 发布 Stable Audio 3.0，加入共享规格；预置 SD 3.5 图像代际较旧",
+      "newPresetIds": [],
+      "modelCount": 1,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/providers/openrouter.json",
+      "name": "openrouter",
+      "source": "https://openrouter.ai/api/v1/models",
+      "finding": "读取官方公共目录，新增当前通用模型；价格按该平台 USD 目录",
+      "newPresetIds": [
+        "anthropic/claude-opus-5.5",
+        "anthropic/claude-sonnet-5.5",
+        "cohere/command-a-plus",
+        "deepseek/deepseek-v4.1-flash",
+        "google/gemini-3.8-flash",
+        "minimax/minimax-m3",
+        "mistralai/mistral-medium-3-5",
+        "moonshotai/kimi-k3",
+        "openai/gpt-6.1-sol",
+        "qwen/qwen3.8-max-prime",
+        "qwen/qwen3.8-omni-flash",
+        "tencent/hy4-preview",
+        "z-ai/glm-5.3",
+        "z-ai/glm-5.3-flash"
+      ],
+      "modelCount": 18,
+      "configurationChanged": true
+    },
+    {
+      "config": "compute/providers/perplexity.json",
+      "name": "perplexity",
+      "source": "https://docs.perplexity.ai/docs/agent-api/migrate-from-sonar/overview",
+      "finding": "官方推动 Sonar 迁移到 Agent API；不是改静态模型名即可完成，保留旧接入并记录迁移事项",
+      "newPresetIds": [],
+      "modelCount": 3,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/coding-plans/zhipu-coding.json",
+      "name": "zhipu-coding",
+      "source": "https://docs.bigmodel.cn/cn/coding-plan/latest-model.md",
+      "finding": "新增 5.3，保留 Flash，套餐仍不提供 FlashX",
+      "newPresetIds": [
+        "glm-5.3"
+      ],
+      "modelCount": 5,
+      "configurationChanged": true
+    },
+    {
+      "config": "compute/coding-plans/dashscope-coding.json",
+      "name": "dashscope-coding",
+      "source": "https://help.aliyun.com/zh/model-studio/coding-plan",
+      "finding": "当前仍提供 qwen3.7-plus 等旧套餐名单，未混用 Token Plan 名单",
+      "newPresetIds": [],
+      "modelCount": 10,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/coding-plans/dashscope-token-plan.json",
+      "name": "dashscope-token-plan",
+      "source": "https://help.aliyun.com/zh/model-studio/token-plan-personal-overview",
+      "finding": "新增当前官方名单中的 DeepSeek/GLM/Image/HappyHorse；音频协议待客户端支持",
+      "newPresetIds": [
+        "deepseek-v4-flash-0731",
+        "deepseek-v4-pro",
+        "deepseek-v4-pro-0813",
+        "deepseek-v4.1-flash",
+        "glm-5.2",
+        "glm-5.3",
+        "happyhorse-1.1-i2v",
+        "happyhorse-1.1-r2v",
+        "happyhorse-1.1-t2v",
+        "qwen-image-3.0-pro",
+        "wan2.7-image",
+        "wan2.7-image-pro"
+      ],
+      "modelCount": 22,
+      "configurationChanged": true
+    },
+    {
+      "config": "compute/coding-plans/volcengine-coding.json",
+      "name": "volcengine-coding",
+      "source": "https://docs.volcengine.com/docs/ark/coding-plan-personal-model-release?lang=zh",
+      "finding": "检索发现 K3 / K2.8 Preview / V4.1 Flash；版本化 API ID 与当前套餐端点配额待确认，未猜测价格/限制",
+      "newPresetIds": [],
+      "modelCount": 14,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/coding-plans/tencent-token.json",
+      "name": "tencent-token",
+      "source": "https://cloud.tencent.com.cn/document/product/1823/130095",
+      "finding": "官方结果支持 tc-code-latest；完整最新套餐清单读取受限",
+      "newPresetIds": [],
+      "modelCount": 10,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/coding-plans/minimax-coding.json",
+      "name": "minimax-coding",
+      "source": "https://platform.minimax.cn/docs/api-reference/text-anthropic-api",
+      "finding": "新增 M3.1 Flash Preview，始终思考，默认 max；不把它放入按量 API Provider",
+      "newPresetIds": [
+        "MiniMax-M3.1-Flash-Preview"
+      ],
+      "modelCount": 11,
+      "configurationChanged": true
+    },
+    {
+      "config": "compute/coding-plans/moonshot-coding.json",
+      "name": "moonshot-coding",
+      "source": "https://platform.kimi.com/docs/guide/kimi-code-cli",
+      "finding": "套餐固定 kimi-for-coding，底层模型由服务端更新；未替换为不保证套餐可用的 K3 ID",
+      "newPresetIds": [],
+      "modelCount": 1,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/coding-plans/baidu-coding.json",
+      "name": "baidu-coding",
+      "source": "https://cloud.baidu.com/doc/qianfan/s/imlg0beiu",
+      "finding": "官方套餐仍列旧代际，保留 qianfan-code-latest；未把自研或按量新模型自动加入套餐",
+      "newPresetIds": [],
+      "modelCount": 10,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/coding-plans/infini-coding.json",
+      "name": "infini-coding",
+      "source": "https://docs.infini-ai.com/",
+      "finding": "官方入口检索可定位，但正文读取失败；未确认最新套餐 ID",
+      "newPresetIds": [],
+      "modelCount": 1,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/coding-plans/moorethread-coding.json",
+      "name": "moorethread-coding",
+      "source": "https://code.mthreads.com/",
+      "finding": "官方页面仅加载壳，搜索结果显示托管 GLM；mt-coder 最新映射无法确认",
+      "newPresetIds": [],
+      "modelCount": 1,
+      "configurationChanged": false
+    },
+    {
+      "config": "compute/coding-plans/kwai-coding.json",
+      "name": "kwai-coding",
+      "source": "https://arxiv.org/abs/2607.05471",
+      "finding": "KAT-Coder V2.5 技术报告为 2026 新模型，套餐 kwai-coder 映射与可调用 ID 未确认",
+      "newPresetIds": [],
+      "modelCount": 1,
+      "configurationChanged": false
+    }
+  ],
+  "validation": {
+    "schemaFilesPassed": 78,
+    "testsPassed": 80,
+    "liveAccountModelCalls": "未执行；缺少逐个供应商测试账号，官方元数据验证不等同真实调用验收"
+  },
+  "coverageLimitations": [
+    "部分供应商仅提供登录后模型清单；没有读取凭据或调用收费模型。",
+    "当前无问芯穹、摩尔线程套餐模型清单读取受限。",
+    "讯飞 X2.5、腾讯 Hy4 原生端点、可灵 3.0 Turbo 等涉及新的协议/端点；共享规格与预置可调用列表有意区分。",
+    "国内 Kimi 当前价格表未返回模型单价，未采用国际价格换算。"
+  ],
+  "reviewFixes": [
+    "收窄 Mistral 和旧 Command A 的 family/pattern，避免旧型号或 A+ 继承错误规格。",
+    "Cohere 新 Rerank 型号只保留共享规格；当前 Compatibility API 未声明 rerank 支持。",
+    "Cohere Command A+ 限定 Compatibility API reasoning_effort 为 none/high；Embedding Provider 只声明兼容接入支持的文本输入。",
+    "Kimi K3 声明拒绝自定义采样参数，删除显式温度/top_p 默认值；共享规格移除接入面的 thinking 回放配置。"
+  ]
+}