@lobehub/chat 1.118.7 → 1.118.8

package/CHANGELOG.md

@@ -2,6 +2,31 @@
 
 # Changelog
 
+### [Version 1.118.8](https://github.com/lobehub/lobe-chat/compare/v1.118.7...v1.118.8)
+
+<sup>Released on **2025-08-30**</sup>
+
+#### 💄 Styles
+
+- **misc**: Update DeepSeek V3.1 & Gemini 2.5 Flash Image Preview models.
+
+<br/>
+
+<details>
+<summary><kbd>Improvements and Fixes</kbd></summary>
+
+#### Styles
+
+- **misc**: Update DeepSeek V3.1 & Gemini 2.5 Flash Image Preview models, closes [#8878](https://github.com/lobehub/lobe-chat/issues/8878) ([5d538a2](https://github.com/lobehub/lobe-chat/commit/5d538a2))
+
+</details>
+
+<div align="right">
+
+[![](https://img.shields.io/badge/-BACK_TO_TOP-151515?style=flat-square)](#readme-top)
+
+</div>
+
 ### [Version 1.118.7](https://github.com/lobehub/lobe-chat/compare/v1.118.6...v1.118.7)
 
 <sup>Released on **2025-08-30**</sup>

package/changelog/v1.json

@@ -1,4 +1,13 @@
 [
+  {
+    "children": {
+      "improvements": [
+        "Update DeepSeek V3.1 & Gemini 2.5 Flash Image Preview models."
+      ]
+    },
+    "date": "2025-08-30",
+    "version": "1.118.8"
+  },
   {
     "children": {
       "improvements": [

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lobehub/chat",
-  "version": "1.118.7",
+  "version": "1.118.8",
   "description": "Lobe Chat - an open-source, high-performance chatbot framework that supports speech synthesis, multimodal, and extensible Function Call plugin system. Supports one-click free deployment of your private ChatGPT/LLM web application.",
   "keywords": [
     "framework",

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

@@ -323,6 +323,27 @@ const aihubmixModels: AIChatModelCard[] = [
     },
     type: 'chat',
   },
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+      vision: true,
+    },
+    contextWindowTokens: 256_000,
+    description:
+      '我们最新最强大的旗舰模型，在自然语言处理、数学计算和推理方面表现卓越 —— 是一款完美的全能型选手。',
+    displayName: 'Grok 4 0709',
+    enabled: true,
+    id: 'grok-4',
+    pricing: {
+      units: [
+        { name: 'textInput', rate: 3.3, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 16.5, strategy: 'fixed', unit: 'millionTokens' },
+      ],
+    },
+    releasedAt: '2025-07-09',
+    type: 'chat',
+  },
   {
     abilities: {
       functionCall: true,
@@ -331,7 +352,6 @@ const aihubmixModels: AIChatModelCard[] = [
     description:
       '旗舰级模型，擅长数据提取、编程和文本摘要等企业级应用，拥有金融、医疗、法律和科学等领域的深厚知识。',
     displayName: 'Grok 3',
-    enabled: true,
     id: 'grok-3',
     pricing: {
       units: [
@@ -352,7 +372,6 @@ const aihubmixModels: AIChatModelCard[] = [
     description:
       '轻量级模型，回话前会先思考。运行快速、智能，适用于不需要深层领域知识的逻辑任务，并能获取原始的思维轨迹。',
     displayName: 'Grok 3 Mini',
-    enabled: true,
     id: 'grok-3-mini',
     pricing: {
       units: [
@@ -535,11 +554,29 @@ const aihubmixModels: AIChatModelCard[] = [
       functionCall: true,
       reasoning: true,
     },
-    contextWindowTokens: 65_536,
+    contextWindowTokens: 131_072,
+    description:
+      'DeepSeek-V3.1 是深度求索全新推出的混合推理模型，支持思考与非思考2种推理模式，较 DeepSeek-R1-0528 思考效率更高。经 Post-Training 优化，Agent 工具使用与智能体任务表现大幅提升。',
+    displayName: 'DeepSeek V3.1',
+    enabled: true,
+    id: 'DeepSeek-V3.1',
+    pricing: {
+      units: [
+        { name: 'textInput', rate: 0.56, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 1.68, strategy: 'fixed', unit: 'millionTokens' },
+      ],
+    },
+    type: 'chat',
+  },
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+    },
+    contextWindowTokens: 131_072,
     description:
       '已升级至最新版本250528；字节火山云开源部署的满血 R1，总参数量 671B，输入最高 64k。目前最稳定，推荐用这个。',
     displayName: 'DeepSeek R1',
-    enabled: true,
     id: 'DeepSeek-R1',
     pricing: {
       units: [
@@ -571,10 +608,9 @@ const aihubmixModels: AIChatModelCard[] = [
     abilities: {
       functionCall: true,
     },
-    contextWindowTokens: 65_536,
+    contextWindowTokens: 131_072,
     description: '字节火山云开源部署目前最稳定，推荐用这个。已经自动升级为最新发布的版本 250324 。',
     displayName: 'DeepSeek V3',
-    enabled: true,
     id: 'DeepSeek-V3',
     pricing: {
       units: [
@@ -657,6 +693,25 @@ const aihubmixModels: AIChatModelCard[] = [
     },
     type: 'chat',
   },
+  {
+    abilities: {
+      imageOutput: true,
+      vision: true,
+    },
+    contextWindowTokens: 32_768 + 8192,
+    description: 'Gemini 2.5 Flash 实验模型，支持图像生成',
+    displayName: 'Gemini 2.5 Flash Image Preview',
+    id: 'gemini-2.5-flash-image-preview',
+    maxOutput: 8192,
+    pricing: {
+      units: [
+        { name: 'textInput', rate: 0.3, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 30, strategy: 'fixed', unit: 'millionTokens' },
+      ],
+    },
+    releasedAt: '2025-08-26',
+    type: 'chat',
+  },
   {
     abilities: {
       functionCall: true,

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

@@ -6,10 +6,10 @@ const deepseekChatModels: AIChatModelCard[] = [
     abilities: {
       functionCall: true,
     },
-    contextWindowTokens: 65_536,
+    contextWindowTokens: 131_072,
     description:
-      '最新模型 DeepSeek-V3 多项评测成绩超越 Qwen2.5-72B 和 Llama-3.1-405B 等开源模型，性能对齐领军闭源模型 GPT-4o 与 Claude-3.5-Sonnet。',
-    displayName: 'DeepSeek V3',
+      'DeepSeek V3.1 是 DeepSeek 最新发布的通用大模型，支持混合推理架构，具备更强的 Agent 能力。',
+    displayName: 'DeepSeek V3.1',
     enabled: true,
     id: 'deepseek-chat',
     maxOutput: 8192,
@@ -21,7 +21,7 @@ const deepseekChatModels: AIChatModelCard[] = [
         { name: 'textOutput', rate: 8, strategy: 'fixed', unit: 'millionTokens' },
       ],
     },
-    releasedAt: '2025-03-24',
+    releasedAt: '2025-08-21',
     type: 'chat',
   },
   {
@@ -29,13 +29,13 @@ const deepseekChatModels: AIChatModelCard[] = [
       functionCall: true,
       reasoning: true,
     },
-    contextWindowTokens: 65_536,
+    contextWindowTokens: 131_072,
     description:
-      'DeepSeek 推出的推理模型。在输出最终回答之前，模型会先输出一段思维链内容，以提升最终答案的准确性。',
-    displayName: 'DeepSeek R1',
+      'DeepSeek V3.1 思考模式。在输出最终回答之前，模型会先输出一段思维链内容，以提升最终答案的准确性。',
+    displayName: 'DeepSeek V3.1 Thinking',
     enabled: true,
     id: 'deepseek-reasoner',
-    maxOutput: 8192,
+    maxOutput: 65_536,
     pricing: {
       currency: 'CNY',
       units: [
@@ -44,7 +44,7 @@ const deepseekChatModels: AIChatModelCard[] = [
         { name: 'textOutput', rate: 16, strategy: 'fixed', unit: 'millionTokens' },
       ],
     },
-    releasedAt: '2025-05-28',
+    releasedAt: '2025-08-21',
     type: 'chat',
   },
 ];

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

@@ -7,22 +7,22 @@ const modelscopeChatModels: AIChatModelCard[] = [
       reasoning: true,
     },
     contextWindowTokens: 131_072,
-    description:
-      'DeepSeek R1 通过利用增加的计算资源和在后训练过程中引入算法优化机制，显著提高了其推理和推断能力的深度。该模型在各种基准评估中表现出色，包括数学、编程和一般逻辑方面。其整体性能现已接近领先模型，如 O3 和 Gemini 2.5 Pro。',
-    displayName: 'DeepSeek-R1-0528',
+    description: 'DeepSeek-V3.1 模型为混合推理架构模型，同时支持思考模式与非思考模式。',
+    displayName: 'DeepSeek-V3.1',
     enabled: true,
-    id: 'deepseek-ai/DeepSeek-R1-0528',
+    id: 'deepseek-ai/DeepSeek-V3.1',
     type: 'chat',
   },
   {
     abilities: {
       functionCall: true,
+      reasoning: true,
     },
     contextWindowTokens: 131_072,
-    description: 'DeepSeek-V3是DeepSeek第三代模型的最新版本，具有强大的推理和对话能力。',
-    displayName: 'DeepSeek-V3',
-    enabled: true,
-    id: 'deepseek-ai/DeepSeek-V3',
+    description:
+      'DeepSeek R1 通过利用增加的计算资源和在后训练过程中引入算法优化机制，显著提高了其推理和推断能力的深度。该模型在各种基准评估中表现出色，包括数学、编程和一般逻辑方面。其整体性能现已接近领先模型，如 O3 和 Gemini 2.5 Pro。',
+    displayName: 'DeepSeek-R1-0528',
+    id: 'deepseek-ai/DeepSeek-R1-0528',
     type: 'chat',
   },
   {
@@ -33,10 +33,19 @@ const modelscopeChatModels: AIChatModelCard[] = [
     contextWindowTokens: 131_072,
     description: 'DeepSeek-R1是DeepSeek最新的推理模型，专注于复杂推理任务。',
     displayName: 'DeepSeek-R1',
-    enabled: true,
     id: 'deepseek-ai/DeepSeek-R1',
     type: 'chat',
   },
+  {
+    abilities: {
+      functionCall: true,
+    },
+    contextWindowTokens: 131_072,
+    description: 'DeepSeek-V3是DeepSeek第三代模型的最新版本，具有强大的推理和对话能力。',
+    displayName: 'DeepSeek-V3',
+    id: 'deepseek-ai/DeepSeek-V3',
+    type: 'chat',
+  },
   {
     abilities: {
       functionCall: true,

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

@@ -2,6 +2,130 @@ import { AIChatModelCard } from '@/types/aiModel';
 
 // https://novita.ai/pricing
 const novitaChatModels: AIChatModelCard[] = [
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+    },
+    contextWindowTokens: 163_840,
+    displayName: 'DeepSeek V3.1',
+    id: 'deepseek/deepseek-v3.1',
+    pricing: {
+      units: [
+        { name: 'textInput', rate: 0.55, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 1.66, strategy: 'fixed', unit: 'millionTokens' },
+      ],
+    },
+    type: 'chat',
+  },
+  {
+    abilities: {
+      functionCall: true,
+    },
+    contextWindowTokens: 262_144,
+    displayName: 'Qwen3 Coder 480B A35B Instruct',
+    id: 'qwen/qwen3-coder-480b-a35b-instruct',
+    pricing: {
+      units: [
+        { name: 'textInput', rate: 0.35, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 1.5, strategy: 'fixed', unit: 'millionTokens' },
+      ],
+    },
+    type: 'chat',
+  },
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+    },
+    contextWindowTokens: 131_072,
+    displayName: 'OpenAI GPT OSS 120B',
+    id: 'openai/gpt-oss-120b',
+    pricing: {
+      units: [
+        { name: 'textInput', rate: 0.1, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 0.5, strategy: 'fixed', unit: 'millionTokens' },
+      ],
+    },
+    type: 'chat',
+  },
+  {
+    abilities: {
+      reasoning: true,
+    },
+    contextWindowTokens: 131_072,
+    displayName: 'OpenAI: GPT OSS 20B',
+    id: 'openai/gpt-oss-20b',
+    pricing: {
+      units: [
+        { name: 'textInput', rate: 0.05, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 0.2, strategy: 'fixed', unit: 'millionTokens' },
+      ],
+    },
+    type: 'chat',
+  },
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+    },
+    contextWindowTokens: 131_072,
+    displayName: 'GLM-4.5',
+    id: 'zai-org/glm-4.5',
+    pricing: {
+      units: [
+        { name: 'textInput', rate: 0.6, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 2.2, strategy: 'fixed', unit: 'millionTokens' },
+      ],
+    },
+    type: 'chat',
+  },
+  {
+    abilities: {
+      functionCall: true,
+    },
+    contextWindowTokens: 262_144,
+    displayName: 'Qwen3 235B A22B Instruct 2507',
+    id: 'qwen/qwen3-235b-a22b-instruct-2507',
+    pricing: {
+      units: [
+        { name: 'textInput', rate: 0.15, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 0.8, strategy: 'fixed', unit: 'millionTokens' },
+      ],
+    },
+    type: 'chat',
+  },
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+    },
+    contextWindowTokens: 131_072,
+    displayName: 'Qwen3 235B A22b Thinking 2507',
+    id: 'qwen/qwen3-235b-a22b-thinking-2507',
+    pricing: {
+      units: [
+        { name: 'textInput', rate: 0.3, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 3, strategy: 'fixed', unit: 'millionTokens' },
+      ],
+    },
+    type: 'chat',
+  },
+  {
+    abilities: {
+      reasoning: true,
+    },
+    contextWindowTokens: 131_072,
+    displayName: 'BaiChuan M2 32B',
+    id: 'baichuan/baichuan-m2-32b',
+    pricing: {
+      units: [
+        { name: 'textInput', rate: 0.07, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 0.07, strategy: 'fixed', unit: 'millionTokens' },
+      ],
+    },
+    type: 'chat',
+  },
   {
     abilities: {
       reasoning: true,
@@ -34,6 +158,9 @@ const novitaChatModels: AIChatModelCard[] = [
     type: 'chat',
   },
   {
+    abilities: {
+      functionCall: true,
+    },
     contextWindowTokens: 120_000,
     displayName: 'ERNIE 4.5 0.3B',
     id: 'baidu/ernie-4.5-0.3b',
@@ -46,13 +173,16 @@ const novitaChatModels: AIChatModelCard[] = [
     type: 'chat',
   },
   {
+    abilities: {
+      functionCall: true,
+    },
     contextWindowTokens: 120_000,
     displayName: 'ERNIE 4.5 21B A3B',
     id: 'baidu/ernie-4.5-21B-a3b',
     pricing: {
       units: [
-        { name: 'textInput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textOutput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textInput', rate: 0.07, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 0.28, strategy: 'fixed', unit: 'millionTokens' },
       ],
     },
     type: 'chat',
@@ -74,6 +204,7 @@ const novitaChatModels: AIChatModelCard[] = [
   },
   {
     abilities: {
+      functionCall: true,
       reasoning: true,
       vision: true,
     },
@@ -82,8 +213,8 @@ const novitaChatModels: AIChatModelCard[] = [
     id: 'baidu/ernie-4.5-vl-28b-a3b',
     pricing: {
       units: [
-        { name: 'textInput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textOutput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textInput', rate: 0.14, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 0.56, strategy: 'fixed', unit: 'millionTokens' },
       ],
     },
     type: 'chat',
@@ -211,6 +342,7 @@ const novitaChatModels: AIChatModelCard[] = [
   {
     abilities: {
       functionCall: true,
+      vision: true,
     },
     contextWindowTokens: 131_072,
     displayName: 'Llama 4 Scout 17B Instruct',
@@ -227,6 +359,7 @@ const novitaChatModels: AIChatModelCard[] = [
   {
     abilities: {
       functionCall: true,
+      vision: true,
     },
     contextWindowTokens: 1_048_576,
     displayName: 'Llama 4 Maverick 17B Instruct',
@@ -234,7 +367,7 @@ const novitaChatModels: AIChatModelCard[] = [
     id: 'meta-llama/llama-4-maverick-17b-128e-instruct-fp8',
     pricing: {
       units: [
-        { name: 'textInput', rate: 0.2, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textInput', rate: 0.17, strategy: 'fixed', unit: 'millionTokens' },
         { name: 'textOutput', rate: 0.85, strategy: 'fixed', unit: 'millionTokens' },
       ],
     },
@@ -286,7 +419,7 @@ const novitaChatModels: AIChatModelCard[] = [
     id: 'google/gemma-3-27b-it',
     pricing: {
       units: [
-        { name: 'textInput', rate: 0.2, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textInput', rate: 0.119, strategy: 'fixed', unit: 'millionTokens' },
         { name: 'textOutput', rate: 0.2, strategy: 'fixed', unit: 'millionTokens' },
       ],
     },
@@ -562,6 +695,9 @@ const novitaChatModels: AIChatModelCard[] = [
     type: 'chat',
   },
   {
+    abilities: {
+      functionCall: true,
+    },
     contextWindowTokens: 8192,
     displayName: 'L3 70B Euryale v2.1',
     id: 'sao10k/l3-70b-euryale-v2.1',
@@ -653,6 +789,9 @@ const novitaChatModels: AIChatModelCard[] = [
     type: 'chat',
   },
   {
+    abilities: {
+      functionCall: true,
+    },
     contextWindowTokens: 8192,
     displayName: 'L31 70B Euryale v2.2',
     id: 'sao10k/l31-70b-euryale-v2.2',
@@ -673,8 +812,8 @@ const novitaChatModels: AIChatModelCard[] = [
     id: 'qwen/qwen2.5-7b-instruct',
     pricing: {
       units: [
-        { name: 'textInput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textOutput', rate: 0, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textInput', rate: 0.07, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 0.07, strategy: 'fixed', unit: 'millionTokens' },
       ],
     },
     type: 'chat',
@@ -688,8 +827,8 @@ const novitaChatModels: AIChatModelCard[] = [
     id: 'thudm/glm-4-32b-0414',
     pricing: {
       units: [
-        { name: 'textInput', rate: 0.24, strategy: 'fixed', unit: 'millionTokens' },
-        { name: 'textOutput', rate: 0.24, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textInput', rate: 0.55, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 1.66, strategy: 'fixed', unit: 'millionTokens' },
       ],
     },
     type: 'chat',

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

@@ -11,6 +11,25 @@ const openrouterChatModels: AIChatModelCard[] = [
     id: 'openrouter/auto',
     type: 'chat',
   },
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+    },
+    contextWindowTokens: 163_840,
+    description:
+      'DeepSeek-V3.1是一款支持128K长上下文和高效模式切换的大型混合推理模型，它在工具调用、代码生成和复杂推理任务上实现了卓越的性能与速度。',
+    displayName: 'DeepSeek V3.1',
+    id: 'deepseek/deepseek-chat-v3.1',
+    pricing: {
+      units: [
+        { name: 'textInput', rate: 0.2, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 0.8, strategy: 'fixed', unit: 'millionTokens' },
+      ],
+    },
+    releasedAt: '2025-08-21',
+    type: 'chat',
+  },
   {
     abilities: {
       imageOutput: true,
@@ -44,6 +63,25 @@ const openrouterChatModels: AIChatModelCard[] = [
     releasedAt: '2025-08-26',
     type: 'chat',
   },
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+    },
+    contextWindowTokens: 163_840,
+    description:
+      'DeepSeek-V3.1是一款支持128K长上下文和高效模式切换的大型混合推理模型，它在工具调用、代码生成和复杂推理任务上实现了卓越的性能与速度。',
+    displayName: 'DeepSeek V3.1',
+    id: 'deepseek/deepseek-chat-v3.1',
+    pricing: {
+      units: [
+        { name: 'textInput', rate: 0.2, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 0.8, strategy: 'fixed', unit: 'millionTokens' },
+      ],
+    },
+    releasedAt: '2025-08-21',
+    type: 'chat',
+  },
   {
     abilities: {
       reasoning: true,

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

@@ -3,6 +3,25 @@ import { AIChatModelCard, AIImageModelCard } from '@/types/aiModel';
 // https://help.aliyun.com/zh/model-studio/models?spm=a2c4g.11186623
 
 const qwenChatModels: AIChatModelCard[] = [
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+    },
+    contextWindowTokens: 131_072,
+    description: 'DeepSeek-V3.1 模型为混合推理架构模型，同时支持思考模式与非思考模式。',
+    displayName: 'DeepSeek V3.1',
+    id: 'deepseek-v3.1',
+    maxOutput: 65_536,
+    pricing: {
+      currency: 'CNY',
+      units: [
+        { name: 'textInput', rate: 4, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 12, strategy: 'fixed', unit: 'millionTokens' },
+      ],
+    },
+    type: 'chat',
+  },
   {
     abilities: {
       search: true,
@@ -11,7 +30,6 @@ const qwenChatModels: AIChatModelCard[] = [
     description:
       '总参数 1T，激活参数 32B。 非思维模型中，在前沿知识、数学和编码方面达到了顶尖水平，更擅长通用 Agent 任务。 针对代理任务进行了精心优化，不仅能回答问题，还能采取行动。 最适用于即兴、通用聊天和代理体验，是一款无需长时间思考的反射级模型。',
     displayName: 'Kimi K2 Instruct',
-    enabled: true,
     id: 'Moonshot-Kimi-K2-Instruct',
     maxOutput: 8192,
     organization: 'Qwen',
@@ -45,7 +63,7 @@ const qwenChatModels: AIChatModelCard[] = [
     pricing: {
       currency: 'CNY',
       units: [
-        { name: 'textInput_cacheRead', rate: 2.4, strategy: 'fixed', unit: 'millionTokens' }, // tokens 32K ~ 128K
+        { name: 'textInput_cacheRead', rate: 6 * 0.2, strategy: 'fixed', unit: 'millionTokens' }, // tokens 32K ~ 128K
         { name: 'textInput', rate: 6, strategy: 'fixed', unit: 'millionTokens' },
         { name: 'textOutput', rate: 24, strategy: 'fixed', unit: 'millionTokens' },
       ],
@@ -70,7 +88,7 @@ const qwenChatModels: AIChatModelCard[] = [
     pricing: {
       currency: 'CNY',
       units: [
-        { name: 'textInput_cacheRead', rate: 0.6, strategy: 'fixed', unit: 'millionTokens' }, // tokens 32K ~ 128K
+        { name: 'textInput_cacheRead', rate: 1.5 * 0.2, strategy: 'fixed', unit: 'millionTokens' }, // tokens 32K ~ 128K
         { name: 'textInput', rate: 1.5, strategy: 'fixed', unit: 'millionTokens' },
         { name: 'textOutput', rate: 6, strategy: 'fixed', unit: 'millionTokens' },
       ],
@@ -155,7 +173,6 @@ const qwenChatModels: AIChatModelCard[] = [
     description:
       '基于Qwen3的思考模式开源模型，相较上一版本（通义千问3-30B-A3B）逻辑能力、通用能力、知识增强及创作能力均有大幅提升，适用于高难度强推理场景。',
     displayName: 'Qwen3 30B A3B Thinking 2507',
-    enabled: true,
     id: 'qwen3-30b-a3b-thinking-2507',
     maxOutput: 32_768,
     organization: 'Qwen',
@@ -180,7 +197,6 @@ const qwenChatModels: AIChatModelCard[] = [
     description:
       '相较上一版本（Qwen3-30B-A3B）中英文和多语言整体通用能力有大幅提升。主观开放类任务专项优化，显著更加符合用户偏好，能够提供更有帮助性的回复。',
     displayName: 'Qwen3 30B A3B Instruct 2507',
-    enabled: true,
     id: 'qwen3-30b-a3b-instruct-2507',
     maxOutput: 32_768,
     organization: 'Qwen',
@@ -466,9 +482,9 @@ const qwenChatModels: AIChatModelCard[] = [
           name: 'textInput_cacheRead',
           strategy: 'tiered',
           tiers: [
-            { rate: 0.15 * 0.4, upTo: 0.128 },
-            { rate: 0.6 * 0.4, upTo: 0.256 },
-            { rate: 1.2 * 0.4, upTo: 'infinity' },
+            { rate: 0.15 * 0.2, upTo: 0.128 },
+            { rate: 0.6 * 0.2, upTo: 0.256 },
+            { rate: 1.2 * 0.2, upTo: 'infinity' },
           ],
           unit: 'millionTokens',
         },
@@ -500,7 +516,7 @@ const qwenChatModels: AIChatModelCard[] = [
     pricing: {
       currency: 'CNY',
       units: [
-        { name: 'textInput_cacheRead', rate: 0.12, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textInput_cacheRead', rate: 0.3 * 0.2, strategy: 'fixed', unit: 'millionTokens' },
         { name: 'textInput', rate: 0.3, strategy: 'fixed', unit: 'millionTokens' },
         { name: 'textOutput', rate: 3, strategy: 'fixed', unit: 'millionTokens' },
       ],
@@ -534,9 +550,9 @@ const qwenChatModels: AIChatModelCard[] = [
         {
           lookup: {
             prices: {
-              '[0, 128_000]': 0.8 * 0.4,
-              '[128_000, 256_000]': 2.4 * 0.4,
-              '[256_000, infinity]': 4.8 * 0.4,
+              '[0, 128_000]': 0.8 * 0.2,
+              '[128_000, 256_000]': 2.4 * 0.2,
+              '[256_000, infinity]': 4.8 * 0.2,
             },
             pricingParams: ['textInputRange'],
           },
@@ -602,7 +618,7 @@ const qwenChatModels: AIChatModelCard[] = [
     pricing: {
       currency: 'CNY',
       units: [
-        { name: 'textInput_cacheRead', rate: 0.96, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textInput_cacheRead', rate: 2.4 * 0.2, strategy: 'fixed', unit: 'millionTokens' },
         { name: 'textInput', rate: 2.4, strategy: 'fixed', unit: 'millionTokens' },
         { name: 'textOutput', rate: 9.6, strategy: 'fixed', unit: 'millionTokens' },
       ],
@@ -695,7 +711,7 @@ const qwenChatModels: AIChatModelCard[] = [
     pricing: {
       currency: 'CNY',
       units: [
-        { name: 'textInput_cacheRead', rate: 0.8 * 0.4, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textInput_cacheRead', rate: 0.8 * 0.2, strategy: 'fixed', unit: 'millionTokens' },
         { name: 'textInput', rate: 0.8, strategy: 'fixed', unit: 'millionTokens' },
         { name: 'textOutput', rate: 2, strategy: 'fixed', unit: 'millionTokens' },
       ],
@@ -719,7 +735,7 @@ const qwenChatModels: AIChatModelCard[] = [
     pricing: {
       currency: 'CNY',
       units: [
-        { name: 'textInput_cacheRead', rate: 1.6 * 0.4, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textInput_cacheRead', rate: 1.6 * 0.2, strategy: 'fixed', unit: 'millionTokens' },
         { name: 'textInput', rate: 1.6, strategy: 'fixed', unit: 'millionTokens' },
         { name: 'textOutput', rate: 4, strategy: 'fixed', unit: 'millionTokens' },
       ],

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

@@ -2,6 +2,45 @@ import { AIChatModelCard, AIImageModelCard } from '@/types/aiModel';
 
 // https://siliconflow.cn/zh-cn/models
 const siliconcloudChatModels: AIChatModelCard[] = [
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+    },
+    contextWindowTokens: 163_840,
+    description:
+      'DeepSeek-V3.1 是由深度求索（DeepSeek AI）发布的混合模式大语言模型，它在前代模型的基础上进行了多方面的重要升级。该模型的一大创新是集成了“思考模式”（Thinking Mode）和“非思考模式”（Non-thinking Mode）于一体，用户可以通过调整聊天模板灵活切换，以适应不同的任务需求。通过专门的训练后优化，V3.1 在工具调用和 Agent 任务方面的性能得到了显著增强，能够更好地支持外部搜索工具和执行多步复杂任务。该模型基于 DeepSeek-V3.1-Base 进行后训练，通过两阶段长文本扩展方法，大幅增加了训练数据量，使其在处理长文档和长篇代码方面表现更佳。作为一个开源模型，DeepSeek-V3.1 在编码、数学和推理等多个基准测试中展现了与顶尖闭源模型相媲美的能力，同时凭借其混合专家（MoE）架构，在保持巨大模型容量的同时，有效降低了推理成本。',
+    displayName: 'DeepSeek V3.1',
+    enabled: true,
+    id: 'deepseek-ai/DeepSeek-V3.1',
+    pricing: {
+      currency: 'CNY',
+      units: [
+        { name: 'textInput', rate: 4, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 12, strategy: 'fixed', unit: 'millionTokens' },
+      ],
+    },
+    type: 'chat',
+  },
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+    },
+    contextWindowTokens: 163_840,
+    description:
+      'DeepSeek-V3.1 是由深度求索（DeepSeek AI）发布的混合模式大语言模型，它在前代模型的基础上进行了多方面的重要升级。该模型的一大创新是集成了“思考模式”（Thinking Mode）和“非思考模式”（Non-thinking Mode）于一体，用户可以通过调整聊天模板灵活切换，以适应不同的任务需求。通过专门的训练后优化，V3.1 在工具调用和 Agent 任务方面的性能得到了显著增强，能够更好地支持外部搜索工具和执行多步复杂任务。该模型基于 DeepSeek-V3.1-Base 进行后训练，通过两阶段长文本扩展方法，大幅增加了训练数据量，使其在处理长文档和长篇代码方面表现更佳。作为一个开源模型，DeepSeek-V3.1 在编码、数学和推理等多个基准测试中展现了与顶尖闭源模型相媲美的能力，同时凭借其混合专家（MoE）架构，在保持巨大模型容量的同时，有效降低了推理成本。',
+    displayName: 'DeepSeek V3.1 (Pro)',
+    id: 'Pro/deepseek-ai/DeepSeek-V3.1',
+    pricing: {
+      currency: 'CNY',
+      units: [
+        { name: 'textInput', rate: 4, strategy: 'fixed', unit: 'millionTokens' },
+        { name: 'textOutput', rate: 12, strategy: 'fixed', unit: 'millionTokens' },
+      ],
+    },
+    type: 'chat',
+  },
   {
     abilities: {
       functionCall: true,
@@ -452,7 +491,6 @@ const siliconcloudChatModels: AIChatModelCard[] = [
     description:
       'Qwen3是一款能力大幅提升的新一代通义千问大模型，在推理、通用、Agent和多语言等多个核心能力上均达到业界领先水平，并支持思考模式切换。',
     displayName: 'Qwen3 8B (Free)',
-    enabled: true,
     id: 'Qwen/Qwen3-8B',
     organization: 'Qwen',
     pricing: {

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

@@ -40,7 +40,6 @@ const vertexaiChatModels: AIChatModelCard[] = [
     description:
       'Gemini 2.5 Pro Preview 是 Google 最先进的思维模型，能够对代码、数学和STEM领域的复杂问题进行推理，以及使用长上下文分析大型数据集、代码库和文档。',
     displayName: 'Gemini 2.5 Pro Preview 05-06',
-    enabled: true,
     id: 'gemini-2.5-pro-preview-05-06',
     maxOutput: 65_536,
     pricing: {
@@ -109,7 +108,6 @@ const vertexaiChatModels: AIChatModelCard[] = [
     contextWindowTokens: 1_048_576 + 65_536,
     description: 'Gemini 2.5 Flash Preview 是 Google 性价比最高的模型，提供全面的功能。',
     displayName: 'Gemini 2.5 Flash Preview 04-17',
-    enabled: true,
     id: 'gemini-2.5-flash-preview-04-17',
     maxOutput: 65_536,
     pricing: {
@@ -153,7 +151,6 @@ const vertexaiChatModels: AIChatModelCard[] = [
     contextWindowTokens: 1_000_000 + 64_000,
     description: 'Gemini 2.5 Flash-Lite 是 Google 最小、性价比最高的模型，专为大规模使用而设计。',
     displayName: 'Gemini 2.5 Flash-Lite',
-    enabled: true,
     id: 'gemini-2.5-flash-lite',
     maxOutput: 64_000,
     pricing: {
@@ -180,7 +177,6 @@ const vertexaiChatModels: AIChatModelCard[] = [
     description:
       'Gemini 2.5 Flash-Lite Preview 是 Google 最小、性价比最高的模型，专为大规模使用而设计。',
     displayName: 'Gemini 2.5 Flash-Lite Preview 06-17',
-    enabled: true,
     id: 'gemini-2.5-flash-lite-preview-06-17',
     maxOutput: 64_000,
     pricing: {

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

@@ -4,6 +4,33 @@ import { AIChatModelCard, AIImageModelCard } from '@/types/aiModel';
 // pricing https://console.volcengine.com/ark/region:ark+cn-beijing/openManagement
 
 const doubaoChatModels: AIChatModelCard[] = [
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+    },
+    config: {
+      deploymentName: 'deepseek-v3-1-250821',
+    },
+    contextWindowTokens: 131_072,
+    description:
+      'DeepSeek-V3.1 是深度求索全新推出的混合推理模型，支持思考与非思考2种推理模式，较 DeepSeek-R1-0528 思考效率更高。经 Post-Training 优化，Agent 工具使用与智能体任务表现大幅提升。支持 128k 上下文窗口，输出长度支持最大 64k tokens。',
+    displayName: 'DeepSeek V3.1',
+    id: 'deepseek-v3.1',
+    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: 12, strategy: 'fixed', unit: 'millionTokens' },
+      ],
+    },
+    settings: {
+      extendParams: ['enableReasoning'],
+    },
+    type: 'chat',
+  },
   {
     abilities: {
       functionCall: true,

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

@@ -11,7 +11,6 @@ const xaiChatModels: AIChatModelCard[] = [
     description:
       '我们很高兴推出 grok-code-fast-1，这是一款快速且经济高效的推理模型，在代理编码方面表现出色。',
     displayName: 'Grok Code Fast 1',
-    enabled: true,
     id: 'grok-code-fast-1',
     pricing: {
       units: [
@@ -20,7 +19,7 @@ const xaiChatModels: AIChatModelCard[] = [
         { name: 'textOutput', rate: 1.5, strategy: 'fixed', unit: 'millionTokens' },
       ],
     },
-    releasedAt: '2025-08-28',
+    releasedAt: '2025-08-27',
     // settings: {
       // reasoning_effort is not supported by grok-code. Specifying reasoning_effort parameter will get an error response.
       // extendParams: ['reasoningEffort'],

package/packages/model-runtime/src/deepseek/index.ts

@@ -1,6 +1,5 @@
-import type { ChatModelCard } from '@/types/llm';
-
 import { ModelProvider } from '../types';
+import { MODEL_LIST_CONFIGS, processModelList } from '../utils/modelParse';
 import { createOpenAICompatibleRuntime } from '../utils/openaiCompatibleFactory';
 
 export interface DeepSeekModelCard {
@@ -18,29 +17,7 @@ export const LobeDeepSeekAI = createOpenAICompatibleRuntime({
     const modelsPage = (await client.models.list()) as any;
     const modelList: DeepSeekModelCard[] = modelsPage.data;
 
-    return modelList
-      .map((model) => {
-        const knownModel = LOBE_DEFAULT_MODEL_LIST.find(
-          (m) => model.id.toLowerCase() === m.id.toLowerCase(),
-        );
-
-        return {
-          contextWindowTokens: knownModel?.contextWindowTokens ?? undefined,
-          displayName: knownModel?.displayName ?? undefined,
-          enabled: knownModel?.enabled || false,
-          functionCall:
-            !model.id.toLowerCase().includes('reasoner') ||
-            knownModel?.abilities?.functionCall ||
-            false,
-          id: model.id,
-          reasoning:
-            model.id.toLowerCase().includes('reasoner') ||
-            knownModel?.abilities?.reasoning ||
-            false,
-          vision: knownModel?.abilities?.vision || false,
-        };
-      })
-      .filter(Boolean) as ChatModelCard[];
+    return processModelList(modelList, MODEL_LIST_CONFIGS.deepseek, 'deepseek');
   },
   provider: ModelProvider.DeepSeek,
 });

package/packages/model-runtime/src/google/index.ts

@@ -134,7 +134,7 @@ export class LobeGoogleAI implements LobeRuntimeAI {
       const thinkingConfig: ThinkingConfig = {
         includeThoughts:
           !!thinkingBudget ||
-          (!thinkingBudget && model && (model.includes('-2.5-') || model.includes('thinking')))
+            (!thinkingBudget && model && (model.includes('-2.5-') || model.includes('thinking')))
             ? true
             : undefined,
         // https://ai.google.dev/gemini-api/docs/thinking#set-budget

package/packages/model-runtime/src/utils/modelParse.ts

@@ -17,8 +17,8 @@ export const MODEL_LIST_CONFIGS = {
     visionKeywords: ['claude'],
   },
   deepseek: {
-    functionCallKeywords: ['v3', 'r1'],
-    reasoningKeywords: ['r1'],
+    functionCallKeywords: ['v3', 'r1', 'deepseek-chat'],
+    reasoningKeywords: ['r1', 'deepseek-reasoner', 'v3.1'],
   },
   google: {
     functionCallKeywords: ['gemini'],
@@ -67,7 +67,7 @@ export const MODEL_LIST_CONFIGS = {
   },
   xai: {
     functionCallKeywords: ['grok'],
-    reasoningKeywords: ['mini', 'grok-4'],
+    reasoningKeywords: ['mini', 'grok-4', 'grok-code-fast'],
     visionKeywords: ['vision', 'grok-4'],
   },
   zeroone: {

package/packages/model-runtime/src/utils/streams/google-ai.ts

@@ -167,9 +167,13 @@ const transformGoogleGenerativeAIStream = (
         if (candidate.finishReason) {
           const chunks: StreamProtocolChunk[] = [imageChunk];
           if (chunk.usageMetadata) {
+            // usageChunks already includes the 'stop' chunk as its first entry when usage exists,
+            // so append usageChunks to avoid sending a duplicate 'stop'.
             chunks.push(...usageChunks);
+          } else {
+            // No usage metadata, we need to send the stop chunk explicitly.
+            chunks.push({ data: candidate.finishReason, id: context?.id, type: 'stop' });
           }
-          chunks.push({ data: candidate.finishReason, id: context?.id, type: 'stop' });
           return chunks;
         }
 

package/packages/model-runtime/src/volcengine/index.ts

@@ -7,6 +7,7 @@ const THINKING_MODELS = [
   'thinking-pro-m',
   'doubao-seed-1-6',
   'doubao-1-5-ui-tars',
+  'deepseek-v3-1',
 ];
 
 export interface VolcengineModelCard {

package/src/config/modelProviders/modelscope.ts

@@ -51,6 +51,7 @@ const ModelScope: ModelProviderCard = {
   modelList: { showModelFetcher: true },
   name: 'ModelScope',
   settings: {
+    disableBrowserRequest: true, // CORS Error
     proxyUrl: {
       placeholder: 'https://api-inference.modelscope.cn/v1',
     },