@lobehub/chat 1.133.2 → 1.133.4

package/packages/model-bank/src/aiModels/anthropic.ts

@@ -22,7 +22,7 @@ const anthropicChatModels: AIChatModelCard[] = [
         { name: 'textInput_cacheWrite', rate: 3.75, strategy: 'fixed', unit: 'millionTokens' },
       ],
     },
-    releasedAt: '2025-09-30',
+    releasedAt: '2025-09-29',
     settings: {
       extendParams: ['disableContextCaching', 'enableReasoning', 'reasoningBudgetToken'],
       searchImpl: 'params',
@@ -107,7 +107,6 @@ const anthropicChatModels: AIChatModelCard[] = [
     description:
       'Claude Sonnet 4 可以产生近乎即时的响应或延长的逐步思考，用户可以清晰地看到这些过程。API 用户还可以对模型思考的时间进行细致的控制',
     displayName: 'Claude Sonnet 4',
-    enabled: true,
     id: 'claude-sonnet-4-20250514',
     maxOutput: 64_000,
     pricing: {
@@ -172,7 +171,7 @@ const anthropicChatModels: AIChatModelCard[] = [
     contextWindowTokens: 200_000,
     description:
       'Claude 3.5 Sonnet 提供了超越 Opus 的能力和比 Sonnet 更快的速度，同时保持与 Sonnet 相同的价格。Sonnet 特别擅长编程、数据科学、视觉处理、代理任务。',
-    displayName: 'Claude 3.5 Sonnet v2',
+    displayName: 'Claude 3.5 Sonnet (New)',
     id: 'claude-3-5-sonnet-20241022',
     maxOutput: 8192,
     pricing: {
@@ -203,7 +202,7 @@ const anthropicChatModels: AIChatModelCard[] = [
     contextWindowTokens: 200_000,
     description:
       'Claude 3.5 Sonnet 提供了超越 Opus 的能力和比 Sonnet 更快的速度，同时保持与 Sonnet 相同的价格。Sonnet 特别擅长编程、数据科学、视觉处理、代理任务。',
-    displayName: 'Claude 3.5 Sonnet',
+    displayName: 'Claude 3.5 Sonnet (Old)',
     id: 'claude-3-5-sonnet-20240620',
     maxOutput: 8192,
     pricing: {
@@ -235,7 +234,6 @@ const anthropicChatModels: AIChatModelCard[] = [
     description:
       'Claude 3.5 Haiku 是 Anthropic 最快的下一代模型。与 Claude 3 Haiku 相比，Claude 3.5 Haiku 在各项技能上都有所提升，并在许多智力基准测试中超越了上一代最大的模型 Claude 3 Opus。',
     displayName: 'Claude 3.5 Haiku',
-    enabled: true,
     id: 'claude-3-5-haiku-20241022',
     maxOutput: 8192,
     pricing: {
@@ -287,33 +285,6 @@ const anthropicChatModels: AIChatModelCard[] = [
     },
     type: 'chat',
   },
-  {
-    abilities: {
-      functionCall: true,
-      vision: true,
-    },
-    contextWindowTokens: 200_000,
-    description:
-      'Claude 3 Sonnet 在智能和速度方面为企业工作负载提供了理想的平衡。它以更低的价格提供最大效用，可靠且适合大规模部署。',
-    displayName: 'Claude 3 Sonnet',
-    id: 'claude-3-sonnet-20240229', // 弃用日期 2025年7月21日
-    maxOutput: 4096,
-    pricing: {
-      units: [
-        { name: 'textInput_cacheRead', rate: 0.3, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textInput', rate: 3, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textOutput', rate: 15, strategy: 'fixed', unit: 'millionTokens' },
-        {
-          lookup: { prices: { '1h': 6, '5m': 3.75 }, pricingParams: ['ttl'] },
-          name: 'textInput_cacheWrite',
-          strategy: 'lookup',
-          unit: 'millionTokens',
-        },
-      ],
-    },
-    releasedAt: '2024-02-29',
-    type: 'chat',
-  },
   {
     abilities: {
       functionCall: true,
@@ -344,38 +315,6 @@ const anthropicChatModels: AIChatModelCard[] = [
     },
     type: 'chat',
   },
-  {
-    contextWindowTokens: 200_000,
-    description:
-      'Claude 2 为企业提供了关键能力的进步，包括业界领先的 200K token 上下文、大幅降低模型幻觉的发生率、系统提示以及一个新的测试功能：工具调用。',
-    displayName: 'Claude 2.1',
-    id: 'claude-2.1', // 弃用日期 2025年7月21日
-    maxOutput: 4096,
-    pricing: {
-      units: [
-        { name: 'textInput', rate: 8, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textOutput', rate: 24, strategy: 'fixed', unit: 'millionTokens' },
-      ],
-    },
-    releasedAt: '2023-11-21',
-    type: 'chat',
-  },
-  {
-    contextWindowTokens: 100_000,
-    description:
-      'Claude 2 为企业提供了关键能力的进步，包括业界领先的 200K token 上下文、大幅降低模型幻觉的发生率、系统提示以及一个新的测试功能：工具调用。',
-    displayName: 'Claude 2.0',
-    id: 'claude-2.0', // 弃用日期 2025年7月21日
-    maxOutput: 4096,
-    pricing: {
-      units: [
-        { name: 'textInput', rate: 8, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textOutput', rate: 24, strategy: 'fixed', unit: 'millionTokens' },
-      ],
-    },
-    releasedAt: '2023-07-11',
-    type: 'chat',
-  },
 ];
 
 export const allModels = [...anthropicChatModels];

package/packages/model-bank/src/aiModels/google.ts

@@ -1,6 +1,11 @@
 import { CHAT_MODEL_IMAGE_GENERATION_PARAMS, ModelParamsSchema } from '../standard-parameters';
 import { AIChatModelCard, AIImageModelCard } from '../types';
 
+/**
+ * gemini implicit caching not extra cost
+ * https://openrouter.ai/docs/features/prompt-caching#implicit-caching
+ */
+
 const googleChatModels: AIChatModelCard[] = [
   {
     abilities: {
@@ -233,6 +238,7 @@ const googleChatModels: AIChatModelCard[] = [
     pricing: {
       units: [
         { name: 'textInput', rate: 0.3, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'imageInput', rate: 0.3, strategy: 'fixed', unit: 'millionTokens' },
         { name: 'textOutput', rate: 2.5, strategy: 'fixed', unit: 'millionTokens' },
         { name: 'imageOutput', rate: 30, strategy: 'fixed', unit: 'millionTokens' },
       ],

package/packages/model-bank/src/aiModels/novita.ts

@@ -30,8 +30,8 @@ const novitaChatModels: AIChatModelCard[] = [
     maxOutput: 32_768,
     pricing: {
       units: [
-        { name: 'textInput', rate: 0.3, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textOutput', rate: 3, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textInput', rate: 0.98, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 3.95, strategy: 'fixed', unit: 'millionTokens' },
       ],
     },
     type: 'chat',

package/packages/model-bank/src/aiModels/openai.ts

@@ -26,8 +26,7 @@ export const openaiChatModels: AIChatModelCard[] = [
       vision: true,
     },
     contextWindowTokens: 400_000,
-    description:
-      'GPT-5 Codex 是一个针对 Codex 或类似环境中的代理编码任务优化的 GPT-5 版本。',
+    description: 'GPT-5 Codex 是一个针对 Codex 或类似环境中的代理编码任务优化的 GPT-5 版本。',
     displayName: 'GPT-5 Codex',
     id: 'gpt-5-codex',
     maxOutput: 128_000,
@@ -1047,25 +1046,6 @@ export const openaiImageModels: AIImageModelCard[] = [
         { name: 'imageInput', rate: 10, strategy: 'fixed', unit: 'millionTokens' },
         { name: 'imageInput_cacheRead', rate: 2.5, strategy: 'fixed', unit: 'millionTokens' },
         { name: 'imageOutput', rate: 40, strategy: 'fixed', unit: 'millionTokens' },
-        {
-          lookup: {
-            prices: {
-              low_1024x1024: 0.011,
-              low_1024x1536: 0.016,
-              low_1536x1024: 0.016,
-              medium_1024x1024: 0.042,
-              medium_1024x1536: 0.063,
-              medium_1536x1024: 0.063,
-              high_1024x1024: 0.167,
-              high_1024x1536: 0.25,
-              high_1536x1024: 0.25,
-            },
-            pricingParams: ['quality', 'size'],
-          },
-          name: 'imageGeneration',
-          strategy: 'lookup',
-          unit: 'image',
-        },
       ],
     },
     resolutions: ['1024x1024', '1024x1536', '1536x1024'],
@@ -1078,6 +1058,10 @@ export const openaiImageModels: AIImageModelCard[] = [
     id: 'dall-e-3',
     parameters: {
       prompt: { default: '' },
+      quality: {
+        default: 'standard',
+        enum: ['standard', 'hd'],
+      },
       size: {
         default: '1024x1024',
         enum: ['1024x1024', '1792x1024', '1024x1792'],
@@ -1151,7 +1135,7 @@ export const openaiRealtimeModels: AIRealtimeModelCard[] = [
     description: '通用实时模型，支持文本与音频的实时输入输出，并支持图像输入。',
     displayName: 'GPT Realtime',
     id: 'gpt-realtime',
-    maxOutput: 4_096,
+    maxOutput: 4096,
     pricing: {
       units: [
         { name: 'audioInput', rate: 32, strategy: 'fixed', unit: 'millionTokens' },

package/packages/model-bank/src/aiModels/qwen.ts

@@ -55,6 +55,27 @@ const qwenChatModels: AIChatModelCard[] = [
     },
     type: 'chat',
   },
+  {
+    abilities: {
+      reasoning: true,
+    },
+    contextWindowTokens: 131_072,
+    description: 'deepseek-v3.2-exp 引入稀疏注意力机制，旨在提升处理长文本时的训练与推理效率，价格低于 deepseek-v3.1。',
+    displayName: 'DeepSeek V3.2 Exp',
+    id: 'deepseek-v3.2-exp',
+    maxOutput: 65_536,
+    pricing: {
+      currency: 'CNY',
+      units: [
+        { name: 'textInput', rate: 2, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 3, strategy: 'fixed', unit: 'millionTokens' },
+      ],
+    },
+    settings: {
+      extendParams: ['enableReasoning', 'reasoningBudgetToken'],
+    },
+    type: 'chat',
+  },
   {
     abilities: {
       reasoning: true,

package/packages/model-bank/src/aiModels/zhipu.ts

@@ -1,6 +1,72 @@
 import { AIChatModelCard, AIImageModelCard } from '../types/aiModel';
 
+// price: https://bigmodel.cn/pricing
+// ref: https://docs.bigmodel.cn/cn/guide/start/model-overview
+
 const zhipuChatModels: AIChatModelCard[] = [
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+      search: true,
+    },
+    contextWindowTokens: 200_000,
+    description:
+      '智谱最新旗舰模型 GLM-4.6 (355B) 在高级编码、长文本处理、推理与智能体能力上全面超越前代，尤其在编程能力上对齐 Claude Sonnet 4，成为国内顶尖的 Coding 模型。',
+    displayName: 'GLM-4.6',
+    enabled: true,
+    id: 'glm-4.6',
+    maxOutput: 128_000,
+    pricing: {
+      currency: 'CNY',
+      units: [
+        {
+          lookup: {
+            prices: {
+              '[0, 32_000]_[0, 200]': 0.4,
+              '[0, 32_000]_[200, infinity]': 0.6,
+              '[32_000, 200_000]': 0.8,
+            },
+            pricingParams: ['textInput', 'textOutput'],
+          },
+          name: 'textInput_cacheRead',
+          strategy: 'lookup',
+          unit: 'millionTokens',
+        },
+        {
+          lookup: {
+            prices: {
+              '[0, 32_000]_[0, 200]': 2,
+              '[0, 32_000]_[200, infinity]': 3,
+              '[32_000, 200_000]': 4,
+            },
+            pricingParams: ['textInput', 'textOutput'],
+          },
+          name: 'textInput',
+          strategy: 'lookup',
+          unit: 'millionTokens',
+        },
+        {
+          lookup: {
+            prices: {
+              '[0, 32_000]_[0, 200]': 8,
+              '[0, 32_000]_[200, infinity]': 14,
+              '[32_000, 200_000]': 16,
+            },
+            pricingParams: ['textInput', 'textOutput'],
+          },
+          name: 'textOutput',
+          strategy: 'lookup',
+          unit: 'millionTokens',
+        },
+      ],
+    },
+    settings: {
+      extendParams: ['enableReasoning'],
+      searchImpl: 'params',
+    },
+    type: 'chat',
+  },
   {
     abilities: {
       functionCall: true,
@@ -18,10 +84,42 @@ const zhipuChatModels: AIChatModelCard[] = [
     pricing: {
       currency: 'CNY',
       units: [
-        // 输入长度 [0, 32]
-        { name: 'textInput_cacheRead', rate: 0.4, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textInput', rate: 2, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textOutput', rate: 6, strategy: 'fixed', unit: 'millionTokens' },
+        {
+          lookup: {
+            prices: {
+              '[0, 32_000]': 0.4,
+              '[32_000, 65_536]': 0.8,
+            },
+            pricingParams: ['textInput'],
+          },
+          name: 'textInput_cacheRead',
+          strategy: 'lookup',
+          unit: 'millionTokens',
+        },
+        {
+          lookup: {
+            prices: {
+              '[0, 32_000]': 2,
+              '[32_000, 65_536]': 4,
+            },
+            pricingParams: ['textInput'],
+          },
+          name: 'textInput',
+          strategy: 'lookup',
+          unit: 'millionTokens',
+        },
+        {
+          lookup: {
+            prices: {
+              '[0, 32_000]': 6,
+              '[32_000, 65_536]': 12,
+            },
+            pricingParams: ['textInput'],
+          },
+          name: 'textOutput',
+          strategy: 'lookup',
+          unit: 'millionTokens',
+        },
       ],
     },
     settings: {
@@ -38,17 +136,52 @@ const zhipuChatModels: AIChatModelCard[] = [
     },
     contextWindowTokens: 128_000,
     description:
-      '智谱最新旗舰模型，支持思考模式切换，综合能力达到开源模型的 SOTA 水平，上下文长度可达128K。',
+      '智谱旗舰模型，支持思考模式切换，综合能力达到开源模型的 SOTA 水平，上下文长度可达128K。',
     displayName: 'GLM-4.5',
-    enabled: true,
     id: 'glm-4.5',
     maxOutput: 32_768,
     pricing: {
       currency: 'CNY',
       units: [
-        { name: 'textInput_cacheRead', rate: 0.8, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textInput', rate: 4, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textOutput', rate: 16, strategy: 'fixed', unit: 'millionTokens' },
+        {
+          lookup: {
+            prices: {
+              '[0, 32_000]_[0, 200]': 0.4,
+              '[0, 32_000]_[200, infinity]': 0.6,
+              '[32_000, 128_000]': 0.8,
+            },
+            pricingParams: ['textInput', 'textOutput'],
+          },
+          name: 'textInput_cacheRead',
+          strategy: 'lookup',
+          unit: 'millionTokens',
+        },
+        {
+          lookup: {
+            prices: {
+              '[0, 32_000]_[0, 200]': 2,
+              '[0, 32_000]_[200, infinity]': 3,
+              '[32_000, 128_000]': 4,
+            },
+            pricingParams: ['textInput', 'textOutput'],
+          },
+          name: 'textInput',
+          strategy: 'lookup',
+          unit: 'millionTokens',
+        },
+        {
+          lookup: {
+            prices: {
+              '[0, 32_000]_[0, 200]': 8,
+              '[0, 32_000]_[200, infinity]': 14,
+              '[32_000, 128_000]': 16,
+            },
+            pricingParams: ['textInput', 'textOutput'],
+          },
+          name: 'textOutput',
+          strategy: 'lookup',
+          unit: 'millionTokens',
+        },
       ],
     },
     settings: {
@@ -71,9 +204,45 @@ const zhipuChatModels: AIChatModelCard[] = [
     pricing: {
       currency: 'CNY',
       units: [
-        { name: 'textInput_cacheRead', rate: 3.2, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textInput', rate: 16, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textOutput', rate: 64, strategy: 'fixed', unit: 'millionTokens' },
+        {
+          lookup: {
+            prices: {
+              '[0, 32_000]': 1.6,
+              '[0, 32_000]_[200, infinity]': 2.4,
+              '[32_000, 128_000]': 3.2,
+            },
+            pricingParams: ['textInput', 'textOutput'],
+          },
+          name: 'textInput_cacheRead',
+          strategy: 'lookup',
+          unit: 'millionTokens',
+        },
+        {
+          lookup: {
+            prices: {
+              '[0, 32_000]_[0, 200]': 8,
+              '[0, 32_000]_[200, infinity]': 12,
+              '[32_000, 128_000]': 16,
+            },
+            pricingParams: ['textInput', 'textOutput'],
+          },
+          name: 'textInput',
+          strategy: 'lookup',
+          unit: 'millionTokens',
+        },
+        {
+          lookup: {
+            prices: {
+              '[0, 32_000]_[0, 200]': 16,
+              '[0, 32_000]_[200, infinity]': 32,
+              '[32_000, 128_000]': 64,
+            },
+            pricingParams: ['textInput', 'textOutput'],
+          },
+          name: 'textOutput',
+          strategy: 'lookup',
+          unit: 'millionTokens',
+        },
       ],
     },
     settings: {
@@ -96,9 +265,43 @@ const zhipuChatModels: AIChatModelCard[] = [
     pricing: {
       currency: 'CNY',
       units: [
-        { name: 'textInput_cacheRead', rate: 0.24, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textInput', rate: 1.2, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textOutput', rate: 8, strategy: 'fixed', unit: 'millionTokens' },
+        {
+          lookup: {
+            prices: {
+              '[0, 32_000]': 0.16,
+              '[32_000, 128_000]': 0.24,
+            },
+            pricingParams: ['textInput'],
+          },
+          name: 'textInput_cacheRead',
+          strategy: 'lookup',
+          unit: 'millionTokens',
+        },
+        {
+          lookup: {
+            prices: {
+              '[0, 32_000]': 0.8,
+              '[32_000, 128_000]': 1.2,
+            },
+            pricingParams: ['textInput'],
+          },
+          name: 'textInput',
+          strategy: 'lookup',
+          unit: 'millionTokens',
+        },
+        {
+          lookup: {
+            prices: {
+              '[0, 32_000]_[0, 200]': 2,
+              '[0, 32_000]_[200, infinity]': 6,
+              '[32_000, 128_000]': 8,
+            },
+            pricingParams: ['textInput', 'textOutput'],
+          },
+          name: 'textOutput',
+          strategy: 'lookup',
+          unit: 'millionTokens',
+        },
       ],
     },
     settings: {
@@ -121,9 +324,43 @@ const zhipuChatModels: AIChatModelCard[] = [
     pricing: {
       currency: 'CNY',
       units: [
-        { name: 'textInput_cacheRead', rate: 1.6, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textInput', rate: 8, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textOutput', rate: 32, strategy: 'fixed', unit: 'millionTokens' },
+        {
+          lookup: {
+            prices: {
+              '[0, 32_000]': 0.8,
+              '[32_000, 128_000]': 1.6,
+            },
+            pricingParams: ['textInput'],
+          },
+          name: 'textInput_cacheRead',
+          strategy: 'lookup',
+          unit: 'millionTokens',
+        },
+        {
+          lookup: {
+            prices: {
+              '[0, 32_000]': 4,
+              '[32_000, 128_000]': 8,
+            },
+            pricingParams: ['textInput'],
+          },
+          name: 'textInput',
+          strategy: 'lookup',
+          unit: 'millionTokens',
+        },
+        {
+          lookup: {
+            prices: {
+              '[0, 32_000]_[0, 200]': 12,
+              '[0, 32_000]_[200, infinity]': 16,
+              '[32_000, 128_000]': 32,
+            },
+            pricingParams: ['textInput', 'textOutput'],
+          },
+          name: 'textOutput',
+          strategy: 'lookup',
+          unit: 'millionTokens',
+        },
       ],
     },
     settings: {
@@ -187,7 +424,6 @@ const zhipuChatModels: AIChatModelCard[] = [
     description:
       'GLM-4.1V-Thinking 系列模型是目前已知10B级别的VLM模型中性能最强的视觉模型，融合了同级别SOTA的各项视觉语言任务，包括视频理解、图片问答、学科解题、OCR文字识别、文档和图表解读、GUI Agent、前端网页Coding、Grounding等，多项任务能力甚至超过8倍参数量的Qwen2.5-VL-72B。通过领先的强化学习技术，模型掌握了通过思维链推理的方式提升回答的准确性和丰富度，从最终效果和可解释性等维度都显著超过传统的非thinking模型。',
     displayName: 'GLM-4.1V-Thinking-Flash',
-    enabled: true,
     id: 'glm-4.1v-thinking-flash',
     maxOutput: 16_384,
     pricing: {
@@ -414,28 +650,6 @@ const zhipuChatModels: AIChatModelCard[] = [
     },
     type: 'chat',
   },
-  {
-    abilities: {
-      functionCall: true,
-      search: true,
-    },
-    contextWindowTokens: 128_000,
-    description:
-      'GLM-4-AllTools 是一个多功能智能体模型，优化以支持复杂指令规划与工具调用，如网络浏览、代码解释和文本生成，适用于多任务执行。',
-    displayName: 'GLM-4-AllTools',
-    id: 'glm-4-alltools',
-    pricing: {
-      currency: 'CNY',
-      units: [
-        { name: 'textInput', rate: 100, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textOutput', rate: 100, strategy: 'fixed', unit: 'millionTokens' },
-      ],
-    },
-    settings: {
-      searchImpl: 'params',
-    },
-    type: 'chat',
-  },
   {
     abilities: {
       functionCall: true,
@@ -479,27 +693,6 @@ const zhipuChatModels: AIChatModelCard[] = [
     },
     type: 'chat',
   },
-  {
-    abilities: {
-      functionCall: true,
-      search: true,
-    },
-    contextWindowTokens: 128_000,
-    description: 'GLM-4 是发布于2024年1月的旧旗舰版本，目前已被更强的 GLM-4-0520 取代。',
-    displayName: 'GLM-4',
-    id: 'glm-4', // 弃用时间 2025年6月30日
-    pricing: {
-      currency: 'CNY',
-      units: [
-        { name: 'textInput', rate: 100, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textOutput', rate: 100, strategy: 'fixed', unit: 'millionTokens' },
-      ],
-    },
-    settings: {
-      searchImpl: 'params',
-    },
-    type: 'chat',
-  },
   {
     abilities: {
       vision: true,