@lobehub/chat 1.50.4 → 1.50.5

package/.env.example

@@ -106,6 +106,7 @@ OPENAI_API_KEY=sk-xxxxxxxxx
 
 ### DeepSeek AI  ####
 
+# DEEPSEEK_PROXY_URL=https://api.deepseek.com/v1
 # DEEPSEEK_API_KEY=xxxxxxxxxxxxxxxxxxxxxxxxxxxxx
 
 ### Qwen AI  ####

package/CHANGELOG.md

@@ -2,6 +2,32 @@
 
 # Changelog
 
+### [Version 1.50.5](https://github.com/lobehub/lobe-chat/compare/v1.50.4...v1.50.5)
+
+<sup>Released on **2025-02-04**</sup>
+
+#### 💄 Styles
+
+- **misc**: Add/Update Aliyun Cloud Models, update GitHub Models.
+
+<br/>
+
+<details>
+<summary><kbd>Improvements and Fixes</kbd></summary>
+
+#### Styles
+
+- **misc**: Add/Update Aliyun Cloud Models, closes [#5613](https://github.com/lobehub/lobe-chat/issues/5613) ([95cd822](https://github.com/lobehub/lobe-chat/commit/95cd822))
+- **misc**: Update GitHub Models, closes [#5683](https://github.com/lobehub/lobe-chat/issues/5683) ([ed4e048](https://github.com/lobehub/lobe-chat/commit/ed4e048))
+
+</details>
+
+<div align="right">
+
+[![](https://img.shields.io/badge/-BACK_TO_TOP-151515?style=flat-square)](#readme-top)
+
+</div>
+
 ### [Version 1.50.4](https://github.com/lobehub/lobe-chat/compare/v1.50.3...v1.50.4)
 
 <sup>Released on **2025-02-04**</sup>

package/changelog/v1.json

@@ -1,4 +1,13 @@
 [
+  {
+    "children": {
+      "improvements": [
+        "Add/Update Aliyun Cloud Models, update GitHub Models."
+      ]
+    },
+    "date": "2025-02-04",
+    "version": "1.50.5"
+  },
   {
     "children": {
       "fixes": [

package/docs/changelog/2025-01-22-new-ai-provider.mdx

@@ -1,8 +1,8 @@
 ---
 title: LobeChat Launches New AI Provider Management System
 description: >-
-  LobeChat has revamped its AI Provider Management System, now supporting custom AI providers and models.
-
+  LobeChat has revamped its AI Provider Management System, now supporting custom
+  AI providers and models.
 tags:
   - LobeChat
   - AI Provider

package/docs/changelog/2025-02-02-deepseek-r1.mdx

@@ -0,0 +1,33 @@
+---
+title: >-
+  LobeChat Integrates DeepSeek R1, Bringing a Revolutionary Chain of Thought Experience
+
+description: >-
+  LobeChat v1.49.12 fully supports the DeepSeek R1 model, providing users with an unprecedented interactive experience in the chain of thought.
+
+tags:
+  - LobeChat
+  - DeepSeek
+  - Chain of Thought
+---
+
+# Perfect Integration of DeepSeek R1 and it's Deep Thinking Experience 🎉
+
+After nearly 10 days of meticulous refinement, LobeChat has fully integrated the DeepSeek R1 model in version v1.49.12, offering users a revolutionary interactive experience in the chain of thought!
+
+## 🚀 Major Updates
+
+- 🤯 **Comprehensive Support for DeepSeek R1**: Now fully integrated in both the Community and Cloud versions ([lobechat.com](https://lobechat.com)).
+- 🧠 **Real-Time Chain of Thought Display**: Transparently presents the AI's reasoning process, making the resolution of complex issues clear and visible.
+- ⚡️ **Deep Thinking Experience**: Utilizing Chain of Thought technology, it provides more insightful AI conversations.
+- 💫 **Intuitive Problem Analysis**: Makes the analysis of complex issues clear and easy to understand.
+
+## 🌟 How to Use
+
+1. Upgrade to LobeChat v1.49.12 or visit [lobechat.com](https://lobechat.com).
+2. Select the DeepSeek R1 model in the settings.
+3. Experience a whole new level of intelligent conversation!
+
+## 📢 Feedback and Support
+
+If you encounter any issues while using the application or have suggestions for new features, feel free to engage with us through GitHub Discussions. Let's work together to create a better LobeChat!

package/docs/changelog/2025-02-02-deepseek-r1.zh-CN.mdx

@@ -0,0 +1,29 @@
+---
+title: LobeChat 重磅集成 DeepSeek R1，带来革命性思维链体验
+description: LobeChat v1.49.12 已完整支持 DeepSeek R1 模型，为用户带来前所未有的思维链交互体验
+tags:
+  - DeepSeek R1
+  - CoT
+  - 思维链
+---
+
+# 完美集成 DeepSeek R1 ，开启思维链新体验
+
+经过近 10 天的精心打磨，LobeChat 已在 v1.49.12 版本中完整集成了 DeepSeek R1 模型，为用户带来革命性的思维链交互体验！
+
+## 🚀 重大更新
+
+- 🤯 **DeepSeek R1 全面支持**: 现已在社区版与 Cloud 版（[lobechat.com](https://lobechat.com)）中完整接入
+- 🧠 **实时思维链展示**: 透明呈现 AI 的推理过程，让复杂问题的解决过程清晰可见
+- ⚡️ **深度思考体验**: 通过 Chain of Thought 技术，带来更具洞察力的 AI 对话
+- 💫 **直观的问题解析**: 让复杂问题的分析过程变得清晰易懂
+
+## 🌟 使用方式
+
+1. 升级到 LobeChat v1.49.12 或访问 [lobechat.com](https://lobechat.com)
+2. 在设置中选择 DeepSeek R1 模型
+3. 开启全新的智能对话体验！
+
+## 📢 反馈与支持
+
+如果您在使用过程中遇到任何问题，或对新功能有任何建议，欢迎通过 GitHub Discussions 与我们交流。让我们一起打造更好的 LobeChat！

package/docs/changelog/index.json

@@ -2,6 +2,12 @@
   "$schema": "https://github.com/lobehub/lobe-chat/blob/main/docs/changelog/schema.json",
   "cloud": [],
   "community": [
+    {
+      "image": "https://github.com/user-attachments/assets/5fe4c373-ebd0-42a9-bdca-0ab7e0a2e747",
+      "id": "2025-02-02-deepseek-r1",
+      "date": "2025-02-02",
+      "versionRange": ["1.47.8", "1.49.12"]
+    },
     {
       "image": "https://github.com/user-attachments/assets/7350f211-61ce-488e-b0e2-f0fcac25caeb",
       "id": "2025-01-22-new-ai-provider",

package/docs/self-hosting/environment-variables/model-provider.mdx

@@ -169,6 +169,13 @@ If you need to use Azure OpenAI to provide model services, you can refer to the
 
 ## DeepSeek AI
 
+### `DEEPSEEK_PROXY_URL`
+
+- Type: Optional
+- Description: If you manually configure the DeepSeek API proxy, you can use this configuration item to override the default DeepSeek API request base URL
+- Default: -
+- Example: `https://xxxx.models.ai.azure.com/v1`
+
 ### `DEEPSEEK_API_KEY`
 
 - Type: Required

package/docs/self-hosting/environment-variables/model-provider.zh-CN.mdx

@@ -167,6 +167,13 @@ LobeChat 在部署时提供了丰富的模型服务商相关的环境变量，
 
 ## DeepSeek AI
 
+### `DEEPSEEK_PROXY_URL`
+
+- 类型：可选
+- 描述：如果您手动配置了 DeepSeek API 代理，可以使用此配置项覆盖默认的 DeepSeek API 请求基础 URL
+- 默认值: -
+- 示例: `https://xxxx.models.ai.azure.com/v1`
+
 ### `DEEPSEEK_API_KEY`
 
 - 类型：必选

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lobehub/chat",
-  "version": "1.50.4",
+  "version": "1.50.5",
   "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/src/config/aiModels/github.ts

@@ -58,7 +58,7 @@ const githubChatModels: AIChatModelCard[] = [
       functionCall: true,
       vision: true,
     },
-    contextWindowTokens: 128_000,
+    contextWindowTokens: 134_144,
     description: '一种经济高效的AI解决方案，适用于多种文本和图像任务。',
     displayName: 'OpenAI GPT-4o mini',
     enabled: true,
@@ -71,11 +71,18 @@ const githubChatModels: AIChatModelCard[] = [
       functionCall: true,
       vision: true,
     },
-    contextWindowTokens: 128_000,
+    contextWindowTokens: 134_144,
     description: 'OpenAI GPT-4系列中最先进的多模态模型，可以处理文本和图像输入。',
     displayName: 'OpenAI GPT-4o',
     enabled: true,
     id: 'gpt-4o',
+    maxOutput: 16_384,
+    type: 'chat',
+  },
+  {
+    contextWindowTokens: 128_000,
+    displayName: 'DeepSeek R1',
+    id: 'DeepSeek-R1',
     maxOutput: 4096,
     type: 'chat',
   },
@@ -145,6 +152,13 @@ const githubChatModels: AIChatModelCard[] = [
     maxOutput: 4096,
     type: 'chat',
   },
+  {
+    contextWindowTokens: 262_144,
+    displayName: 'Codestral',
+    id: 'Codestral-2501',
+    maxOutput: 4096,
+    type: 'chat',
+  },
   {
     abilities: {
       vision: true,
@@ -210,6 +224,20 @@ const githubChatModels: AIChatModelCard[] = [
     maxOutput: 4096,
     type: 'chat',
   },
+  {
+    contextWindowTokens: 16_384,
+    displayName: 'Phi 4',
+    id: 'Phi-4',
+    maxOutput: 16_384,
+    type: 'chat',
+  },
+  {
+    contextWindowTokens: 131_072,
+    displayName: 'Phi 3.5 MoE',
+    id: 'Phi-3.5-MoE-instruct',
+    maxOutput: 4096,
+    type: 'chat',
+  },
   {
     contextWindowTokens: 131_072,
     description: 'Phi-3-mini模型的更新版。',

package/src/config/aiModels/qwen.ts

@@ -5,11 +5,12 @@ const qwenChatModels: AIChatModelCard[] = [
     abilities: {
       functionCall: true,
     },
-    contextWindowTokens: 131_072,
+    contextWindowTokens: 1_000_000,
     description: '通义千问超大规模语言模型，支持中文、英文等不同语言输入。',
     displayName: 'Qwen Turbo',
     enabled: true,
     id: 'qwen-turbo-latest',
+    maxOutput: 8192,
     pricing: {
       currency: 'CNY',
       input: 0.3,
@@ -26,6 +27,7 @@ const qwenChatModels: AIChatModelCard[] = [
     displayName: 'Qwen Plus',
     enabled: true,
     id: 'qwen-plus-latest',
+    maxOutput: 8192,
     pricing: {
       currency: 'CNY',
       input: 0.8,
@@ -43,6 +45,7 @@ const qwenChatModels: AIChatModelCard[] = [
     displayName: 'Qwen Max',
     enabled: true,
     id: 'qwen-max-latest',
+    maxOutput: 8192,
     pricing: {
       currency: 'CNY',
       input: 20,
@@ -56,6 +59,7 @@ const qwenChatModels: AIChatModelCard[] = [
       '通义千问超大规模语言模型，支持长文本上下文，以及基于长文档、多文档等多个场景的对话功能。',
     displayName: 'Qwen Long',
     id: 'qwen-long',
+    maxOutput: 6000,
     pricing: {
       currency: 'CNY',
       input: 0.5,
@@ -73,10 +77,11 @@ const qwenChatModels: AIChatModelCard[] = [
     displayName: 'Qwen VL Plus',
     enabled: true,
     id: 'qwen-vl-plus-latest',
+    maxOutput: 2048,
     pricing: {
       currency: 'CNY',
-      input: 8,
-      output: 8,
+      input: 1.5,
+      output: 4.5,
     },
     type: 'chat',
   },
@@ -84,16 +89,34 @@ const qwenChatModels: AIChatModelCard[] = [
     abilities: {
       vision: true,
     },
-    contextWindowTokens: 32_000,
+    contextWindowTokens: 32_768,
     description:
       '通义千问超大规模视觉语言模型。相比增强版，再次提升视觉推理能力和指令遵循能力，提供更高的视觉感知和认知水平。',
     displayName: 'Qwen VL Max',
     enabled: true,
     id: 'qwen-vl-max-latest',
+    maxOutput: 2048,
     pricing: {
       currency: 'CNY',
-      input: 20,
-      output: 20,
+      input: 3,
+      output: 9,
+    },
+    type: 'chat',
+  },
+  {
+    abilities: {
+      vision: true,
+    },
+    contextWindowTokens: 34_096,
+    description:
+      '通义千问OCR是文字提取专有模型，专注于文档、表格、试题、手写体文字等类型图像的文字提取能力。它能够识别多种文字，目前支持的语言有：汉语、英语、法语、日语、韩语、德语、俄语、意大利语、越南语、阿拉伯语。',
+    displayName: 'Qwen VL OCR',
+    id: 'qwen-vl-ocr-latest',
+    maxOutput: 4096,
+    pricing: {
+      currency: 'CNY',
+      input: 5,
+      output: 5,
     },
     type: 'chat',
   },
@@ -102,6 +125,7 @@ const qwenChatModels: AIChatModelCard[] = [
     description: '通义千问数学模型是专门用于数学解题的语言模型。',
     displayName: 'Qwen Math Turbo',
     id: 'qwen-math-turbo-latest',
+    maxOutput: 3072,
     pricing: {
       currency: 'CNY',
       input: 2,
@@ -114,6 +138,7 @@ const qwenChatModels: AIChatModelCard[] = [
     description: '通义千问数学模型是专门用于数学解题的语言模型。',
     displayName: 'Qwen Math Plus',
     id: 'qwen-math-plus-latest',
+    maxOutput: 3072,
     pricing: {
       currency: 'CNY',
       input: 4,
@@ -126,6 +151,7 @@ const qwenChatModels: AIChatModelCard[] = [
     description: '通义千问代码模型。',
     displayName: 'Qwen Coder Turbo',
     id: 'qwen-coder-turbo-latest',
+    maxOutput: 8192,
     pricing: {
       currency: 'CNY',
       input: 2,
@@ -138,6 +164,7 @@ const qwenChatModels: AIChatModelCard[] = [
     description: '通义千问代码模型。',
     displayName: 'Qwen Coder Plus',
     id: 'qwen-coder-plus-latest',
+    maxOutput: 8192,
     pricing: {
       currency: 'CNY',
       input: 3.5,
@@ -146,10 +173,14 @@ const qwenChatModels: AIChatModelCard[] = [
     type: 'chat',
   },
   {
+    abilities: {
+      functionCall: true,
+    },
     contextWindowTokens: 32_768,
     description: 'QwQ模型是由 Qwen 团队开发的实验性研究模型，专注于增强 AI 推理能力。',
     displayName: 'QwQ 32B Preview',
     id: 'qwq-32b-preview',
+    maxOutput: 16_384,
     pricing: {
       currency: 'CNY',
       input: 3.5,
@@ -166,6 +197,7 @@ const qwenChatModels: AIChatModelCard[] = [
     description: 'QVQ模型是由 Qwen 团队开发的实验性研究模型，专注于提升视觉推理能力，尤其在数学推理领域。',
     displayName: 'QVQ 72B Preview',
     id: 'qvq-72b-preview',
+    maxOutput: 16_384,
     pricing: {
       currency: 'CNY',
       input: 12,
@@ -182,10 +214,11 @@ const qwenChatModels: AIChatModelCard[] = [
     description: '通义千问2.5对外开源的7B规模的模型。',
     displayName: 'Qwen2.5 7B',
     id: 'qwen2.5-7b-instruct',
+    maxOutput: 8192,
     pricing: {
       currency: 'CNY',
-      input: 1,
-      output: 2,
+      input: 0.5,
+      output: 1,
     },
     type: 'chat',
   },
@@ -197,10 +230,11 @@ const qwenChatModels: AIChatModelCard[] = [
     description: '通义千问2.5对外开源的14B规模的模型。',
     displayName: 'Qwen2.5 14B',
     id: 'qwen2.5-14b-instruct',
+    maxOutput: 8192,
     pricing: {
       currency: 'CNY',
-      input: 2,
-      output: 6,
+      input: 1,
+      output: 3,
     },
     type: 'chat',
   },
@@ -212,6 +246,7 @@ const qwenChatModels: AIChatModelCard[] = [
     description: '通义千问2.5对外开源的32B规模的模型。',
     displayName: 'Qwen2.5 32B',
     id: 'qwen2.5-32b-instruct',
+    maxOutput: 8192,
     pricing: {
       currency: 'CNY',
       input: 3.5,
@@ -227,6 +262,7 @@ const qwenChatModels: AIChatModelCard[] = [
     description: '通义千问2.5对外开源的72B规模的模型。',
     displayName: 'Qwen2.5 72B',
     id: 'qwen2.5-72b-instruct',
+    maxOutput: 8192,
     pricing: {
       currency: 'CNY',
       input: 4,
@@ -234,11 +270,29 @@ const qwenChatModels: AIChatModelCard[] = [
     },
     type: 'chat',
   },
+  {
+    abilities: {
+      functionCall: true,
+    },
+    contextWindowTokens: 1_000_000,
+    description: '通义千问2.5对外开源的72B规模的模型。',
+    displayName: 'Qwen2.5 14B 1M',
+    id: 'qwen2.5-14b-instruct-1m',
+    maxOutput: 8192,
+    pricing: {
+      currency: 'CNY',
+      input: 1,
+      output: 3,
+    },
+    releasedAt: '2025-01-27',
+    type: 'chat',
+  },
   {
     contextWindowTokens: 4096,
     description: 'Qwen-Math 模型具有强大的数学解题能力。',
     displayName: 'Qwen2.5 Math 7B',
     id: 'qwen2.5-math-7b-instruct',
+    maxOutput: 3072,
     pricing: {
       currency: 'CNY',
       input: 1,
@@ -251,6 +305,7 @@ const qwenChatModels: AIChatModelCard[] = [
     description: 'Qwen-Math 模型具有强大的数学解题能力。',
     displayName: 'Qwen2.5 Math 72B',
     id: 'qwen2.5-math-72b-instruct',
+    maxOutput: 3072,
     pricing: {
       currency: 'CNY',
       input: 4,
@@ -263,6 +318,7 @@ const qwenChatModels: AIChatModelCard[] = [
     description: '通义千问代码模型开源版。',
     displayName: 'Qwen2.5 Coder 7B',
     id: 'qwen2.5-coder-7b-instruct',
+    maxOutput: 8192,
     pricing: {
       currency: 'CNY',
       input: 1,
@@ -275,6 +331,7 @@ const qwenChatModels: AIChatModelCard[] = [
     description: '通义千问代码模型开源版。',
     displayName: 'Qwen2.5 Coder 32B',
     id: 'qwen2.5-coder-32b-instruct',
+    maxOutput: 8192,
     pricing: {
       currency: 'CNY',
       input: 3.5,
@@ -312,6 +369,78 @@ const qwenChatModels: AIChatModelCard[] = [
     },
     type: 'chat',
   },
+  {
+    abilities: {
+      vision: true,
+    },
+    contextWindowTokens: 131_072,
+    description: 
+      '指令跟随、数学、解题、代码整体提升，万物识别能力提升，支持多样格式直接精准定位视觉元素，支持对长视频文件（最长10分钟）进行理解和秒级别的事件时刻定位，能理解时间先后和快慢，基于解析和定位能力支持操控OS或Mobile的Agent，关键信息抽取能力和Json格式输出能力强，此版本为72B版本，本系列能力最强的版本。',
+    displayName: 'Qwen2.5 VL 72B',
+    id: 'qwen2.5-vl-72b-instruct',
+    maxOutput: 2048,
+    pricing: {
+      currency: 'CNY',
+      input: 16,
+      output: 48,
+    },
+    releasedAt: '2025-01-27',
+    type: 'chat',
+  },
+  {
+    abilities: {
+      vision: true,
+    },
+    contextWindowTokens: 131_072,
+    description: 
+      '指令跟随、数学、解题、代码整体提升，万物识别能力提升，支持多样格式直接精准定位视觉元素，支持对长视频文件（最长10分钟）进行理解和秒级别的事件时刻定位，能理解时间先后和快慢，基于解析和定位能力支持操控OS或Mobile的Agent，关键信息抽取能力和Json格式输出能力强，此版本为72B版本，本系列能力最强的版本。',
+    displayName: 'Qwen2.5 VL 7B',
+    id: 'qwen2.5-vl-7b-instruct',
+    maxOutput: 2048,
+    pricing: {
+      currency: 'CNY',
+      input: 2,
+      output: 5,
+    },
+    releasedAt: '2025-01-27',
+    type: 'chat',
+  },
+  {
+    abilities: {
+      reasoning: true,
+    },
+    contextWindowTokens: 65_536,
+    description: 
+      'DeepSeek-R1 在后训练阶段大规模使用了强化学习技术，在仅有极少标注数据的情况下，极大提升了模型推理能力，尤其在数学、代码、自然语言推理等任务上。',
+    displayName: 'DeepSeek R1',
+    id: 'deepseek-r1',
+    maxOutput: 8192,
+    pricing: {
+      currency: 'CNY',
+      input: 0,
+      output: 0,
+    },
+    releasedAt: '2025-01-27',
+    type: 'chat',
+  },
+  {
+    abilities: {
+      functionCall: true,
+    },
+    contextWindowTokens: 65_536,
+    description: 
+      'DeepSeek-V3 为自研 MoE 模型，671B 参数，激活 37B，在 14.8T token 上进行了预训练，在长文本、代码、数学、百科、中文能力上表现优秀。',
+    displayName: 'DeepSeek V3',
+    id: 'deepseek-v3',
+    maxOutput: 8192,
+    pricing: {
+      currency: 'CNY',
+      input: 0,
+      output: 0,
+    },
+    releasedAt: '2025-01-27',
+    type: 'chat',
+  },
 ];
 
 export const allModels = [...qwenChatModels];

package/src/config/modelProviders/github.ts

@@ -37,7 +37,7 @@ const Github: ModelProviderCard = {
       vision: true,
     },
     {
-      contextWindowTokens: 128_000,
+      contextWindowTokens: 134_144,
       description: '一种经济高效的AI解决方案，适用于多种文本和图像任务。',
       displayName: 'OpenAI GPT-4o mini',
       enabled: true,
@@ -47,15 +47,21 @@ const Github: ModelProviderCard = {
       vision: true,
     },
     {
-      contextWindowTokens: 128_000,
+      contextWindowTokens: 134_144,
       description: 'OpenAI GPT-4系列中最先进的多模态模型，可以处理文本和图像输入。',
       displayName: 'OpenAI GPT-4o',
       enabled: true,
       functionCall: true,
       id: 'gpt-4o',
-      maxOutput: 4096,
+      maxOutput: 16_384,
       vision: true,
     },
+    {
+      contextWindowTokens: 128_000,
+      displayName: 'DeepSeek R1',
+      id: 'DeepSeek-R1',
+      maxOutput: 4096,
+    },
     {
       contextWindowTokens: 262_144,
       description:
@@ -112,6 +118,12 @@ const Github: ModelProviderCard = {
       id: 'mistral-large',
       maxOutput: 4096,
     },
+    {
+      contextWindowTokens: 262_144,
+      displayName: 'Codestral',
+      id: 'Codestral-2501',
+      maxOutput: 4096,
+    },
     {
       contextWindowTokens: 131_072,
       description: '在高分辨率图像上表现出色的图像推理能力，适用于视觉理解应用。',
@@ -166,6 +178,18 @@ const Github: ModelProviderCard = {
       id: 'meta-llama-3-70b-instruct',
       maxOutput: 4096,
     },
+    {
+      contextWindowTokens: 16_384,
+      displayName: 'Phi 4',
+      id: 'Phi-4',
+      maxOutput: 16_384,
+    },
+    {
+      contextWindowTokens: 131_072,
+      displayName: 'Phi 3.5 MoE',
+      id: 'Phi-3.5-MoE-instruct',
+      maxOutput: 4096,
+    },
     {
       contextWindowTokens: 131_072,
       description: 'Phi-3-mini模型的更新版。',

package/src/config/modelProviders/qwen.ts

@@ -4,7 +4,7 @@ import { ModelProviderCard } from '@/types/llm';
 const Qwen: ModelProviderCard = {
   chatModels: [
     {
-      contextWindowTokens: 131_072,
+      contextWindowTokens: 1_000_000,
       description: '通义千问超大规模语言模型，支持中文、英文等不同语言输入。',
       displayName: 'Qwen Turbo',
       enabled: true,
@@ -64,13 +64,13 @@ const Qwen: ModelProviderCard = {
       id: 'qwen-vl-plus-latest',
       pricing: {
         currency: 'CNY',
-        input: 8,
-        output: 8,
+        input: 1.5,
+        output: 4.5,
       },
       vision: true,
     },
     {
-      contextWindowTokens: 32_000,
+      contextWindowTokens: 32_768,
       description:
         '通义千问超大规模视觉语言模型。相比增强版，再次提升视觉推理能力和指令遵循能力，提供更高的视觉感知和认知水平。',
       displayName: 'Qwen VL Max',
@@ -78,8 +78,21 @@ const Qwen: ModelProviderCard = {
       id: 'qwen-vl-max-latest',
       pricing: {
         currency: 'CNY',
-        input: 20,
-        output: 20,
+        input: 3,
+        output: 9,
+      },
+      vision: true,
+    },
+    {
+      contextWindowTokens: 34_096,
+      description:
+        '通义千问OCR是文字提取专有模型，专注于文档、表格、试题、手写体文字等类型图像的文字提取能力。它能够识别多种文字，目前支持的语言有：汉语、英语、法语、日语、韩语、德语、俄语、意大利语、越南语、阿拉伯语。',
+      displayName: 'Qwen VL OCR',
+      id: 'qwen-vl-ocr-latest',
+      pricing: {
+        currency: 'CNY',
+        input: 5,
+        output: 5,
       },
       vision: true,
     },
@@ -134,9 +147,22 @@ const Qwen: ModelProviderCard = {
       id: 'qwq-32b-preview',
       pricing: {
         currency: 'CNY',
-        input: 0,
-        output: 0,
+        input: 3.5,
+        output: 7,
+      },
+    },
+    {
+      contextWindowTokens: 32_768,
+      description: 'QVQ模型是由 Qwen 团队开发的实验性研究模型，专注于提升视觉推理能力，尤其在数学推理领域。',
+      displayName: 'QVQ 72B Preview',
+      id: 'qvq-72b-preview',
+      pricing: {
+        currency: 'CNY',
+        input: 12,
+        output: 36,
       },
+      releasedAt: '2024-12-25',
+      vision: true,
     },
     {
       contextWindowTokens: 131_072,
@@ -146,8 +172,8 @@ const Qwen: ModelProviderCard = {
       id: 'qwen2.5-7b-instruct',
       pricing: {
         currency: 'CNY',
-        input: 1,
-        output: 2,
+        input: 0.5,
+        output: 1,
       },
     },
     {
@@ -158,8 +184,8 @@ const Qwen: ModelProviderCard = {
       id: 'qwen2.5-14b-instruct',
       pricing: {
         currency: 'CNY',
-        input: 2,
-        output: 6,
+        input: 1,
+        output: 3,
       },
     },
     {
@@ -186,6 +212,18 @@ const Qwen: ModelProviderCard = {
         output: 12,
       },
     },
+    {
+      contextWindowTokens: 1_000_000,
+      description: '通义千问2.5对外开源的72B规模的模型。',
+      displayName: 'Qwen2.5 14B 1M',
+      functionCall: true,
+      id: 'qwen2.5-14b-instruct-1m',
+      pricing: {
+        currency: 'CNY',
+        input: 1,
+        output: 3,
+      },
+    },
     {
       contextWindowTokens: 4096,
       description: 'Qwen-Math 模型具有强大的数学解题能力。',
@@ -254,6 +292,46 @@ const Qwen: ModelProviderCard = {
       },
       vision: true,
     },
+    {
+      contextWindowTokens: 128_000,
+      description: 
+        '指令跟随、数学、解题、代码整体提升，万物识别能力提升，支持多样格式直接精准定位视觉元素，支持对长视频文件（最长10分钟）进行理解和秒级别的事件时刻定位，能理解时间先后和快慢，基于解析和定位能力支持操控OS或Mobile的Agent，关键信息抽取能力和Json格式输出能力强，此版本为72B版本，本系列能力最强的版本。',
+      displayName: 'Qwen2.5 VL 72B',
+      id: 'qwen2.5-vl-72b-instruct',
+      pricing: {
+        currency: 'CNY',
+        input: 16,
+        output: 48,
+      },
+      releasedAt: '2025-01-26',
+      vision: true,
+    },
+    {
+      contextWindowTokens: 65_536,
+      description: 
+        'DeepSeek-R1 在后训练阶段大规模使用了强化学习技术，在仅有极少标注数据的情况下，极大提升了模型推理能力，尤其在数学、代码、自然语言推理等任务上。',
+      displayName: 'DeepSeek R1',
+      id: 'deepseek-r1',
+      pricing: {
+        currency: 'CNY',
+        input: 0,
+        output: 0,
+      },
+      releasedAt: '2025-01-27',
+    },
+    {
+      contextWindowTokens: 65_536,
+      description: 
+        'DeepSeek-V3 为自研 MoE 模型，671B 参数，激活 37B，在 14.8T token 上进行了预训练，在长文本、代码、数学、百科、中文能力上表现优秀。',
+      displayName: 'DeepSeek V3',
+      id: 'deepseek-v3',
+      pricing: {
+        currency: 'CNY',
+        input: 0,
+        output: 0,
+      },
+      releasedAt: '2025-01-27',
+    },
   ],
   checkModel: 'qwen-turbo-latest',
   description: