💄 style: Add Qwen QwQ model (lobehub#6783)

* Update groq.ts * Update siliconcloud.ts * Update novita.ts * Update google.ts
bentwnghk · Mar 7, 2025 · 3d3c2ce · 3d3c2ce
1 parent 7f82512
commit 3d3c2ce
Show file tree

Hide file tree

Showing 4 changed files with 68 additions and 3 deletions.
diff --git a/src/config/aiModels/google.ts b/src/config/aiModels/google.ts
@@ -118,7 +118,6 @@ const googleChatModels: AIChatModelCard[] = [
     description:
       'Gemini 2.0 Flash Thinking Exp 是 Google 的实验性多模态推理AI模型，能对复杂问题进行推理，拥有新的思维能力。',
     displayName: 'Gemini 2.0 Flash Thinking Experimental',
-    enabled: true,
     id: 'gemini-2.0-flash-thinking-exp',
     maxOutput: 65_536,
     pricing: {

diff --git a/src/config/aiModels/groq.ts b/src/config/aiModels/groq.ts
@@ -4,6 +4,20 @@ import { AIChatModelCard } from '@/types/aiModel';
 // https://console.groq.com/docs/models
 
 const groqChatModels: AIChatModelCard[] = [
+  {
+    abilities: {
+      functionCall: true,
+      reasoning: true,
+    },
+    contextWindowTokens: 131_072,
+    displayName: 'Qwen QwQ 32B',
+    id: 'qwen-qwq-32b',
+    pricing: {
+      input: 0.29,
+      output: 0.39,
+    },
+    type: 'chat',
+  },
   {
     abilities: {
       functionCall: true,

diff --git a/src/config/aiModels/novita.ts b/src/config/aiModels/novita.ts
@@ -150,6 +150,29 @@ const novitaChatModels: AIChatModelCard[] = [
     },
     type: 'chat',
   },
+  {
+    contextWindowTokens: 64_000,
+    displayName: 'Deepseek V3 Turbo',
+    id: 'deepseek/deepseek-v3-turbo',
+    pricing: {
+      input: 0.4,
+      output: 1.3,
+    },
+    type: 'chat',
+  },
+  {
+    abilities: {
+      reasoning: true,
+    },
+    contextWindowTokens: 64_000,
+    displayName: 'Deepseek R1 Turbo',
+    id: 'deepseek/deepseek-r1-turbo',
+    pricing: {
+      input: 0.7,
+      output: 2.5,
+    },
+    type: 'chat',
+  },
   {
     abilities: {
       reasoning: true,
@@ -378,6 +401,19 @@ const novitaChatModels: AIChatModelCard[] = [
     },
     type: 'chat',
   },
+  {
+    abilities: {
+      reasoning: true,
+    },
+    contextWindowTokens: 32_768,
+    displayName: 'QwQ 32B',
+    id: 'qwen/qwq-32b',
+    pricing: {
+      input: 0.18,
+      output: 0.2,
+    },
+    type: 'chat',
+  },
 ];
 
 export const allModels = [...novitaChatModels];

diff --git a/src/config/aiModels/siliconcloud.ts b/src/config/aiModels/siliconcloud.ts
@@ -226,9 +226,25 @@ const siliconcloudChatModels: AIChatModelCard[] = [
     },
     contextWindowTokens: 32_768,
     description:
-      'QwQ-32B-Preview是Qwen 最新的实验性研究模型，专注于提升AI推理能力。通过探索语言混合、递归推理等复杂机制，主要优势包括强大的推理分析能力、数学和编程能力。与此同时，也存在语言切换问题、推理循环、安全性考虑、其他能力方面的差异。',
-    displayName: 'QwQ 32B Preview',
+      'QwQ 是 Qwen 系列的推理模型。与传统的指令调优模型相比，QwQ 具备思考和推理能力，能够在下游任务中实现显著增强的性能，尤其是在解决困难问题方面。QwQ-32B 是中型推理模型，能够在与最先进的推理模型（如 DeepSeek-R1、o1-mini）的对比中取得有竞争力的性能。该模型采用 RoPE、SwiGLU、RMSNorm 和 Attention QKV bias 等技术，具有 64 层网络结构和 40 个 Q 注意力头（GQA 架构中 KV 为 8 个）。',
+    displayName: 'QwQ 32B',
     enabled: true,
+    id: 'Qwen/QwQ-32B',
+    pricing: {
+      currency: 'CNY',
+      input: 1,
+      output: 4,
+    },
+    type: 'chat',
+  },
+  {
+    abilities: {
+      reasoning: true,
+    },
+    contextWindowTokens: 32_768,
+    description:
+      'QwQ-32B-Preview 是 Qwen 最新的实验性研究模型，专注于提升AI推理能力。通过探索语言混合、递归推理等复杂机制，主要优势包括强大的推理分析能力、数学和编程能力。与此同时，也存在语言切换问题、推理循环、安全性考虑、其他能力方面的差异。',
+    displayName: 'QwQ 32B Preview',
     id: 'Qwen/QwQ-32B-Preview',
     pricing: {
       currency: 'CNY',