references/lobe-chat
1
0
Fork 0
mirror of https://github.com/lobehub/lobe-chat.git synced 2026-06-15 04:00:09 +00:00
Code Issues Packages Projects Releases Wiki Activity

💄 style: update model list, add reasoning tag (#5696)

Browse Source 
* 💄 style: update model list, add reasoning tag

* 💄 style: add `internlm3-latest` for InternLM

* 💄 style: add missing reasoning tag for QvQ

* 💄 style: fix deepseek-r1 reasoning tag missing

* 🐛 fix: fix ci error
This commit is contained in:
Zhijie He
2025-02-06 00:46:50 +08:00
committed by GitHub
parent 283157a84b
commit dedd7840d6
15 changed files with 218 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/config/aiModels/ai360.ts
+22 -2
View File
@@ -2,6 +2,26 @@ import { AIChatModelCard } from '@/types/aiModel';
const ai360ChatModels: AIChatModelCard[] = [
{
abilities: {
reasoning: true,
},
contextWindowTokens: 8000,
description:
'360zhinao2-o1 使用树搜索构建思维链，并引入了反思机制，使用强化学习训练，模型具备自我反思与纠错的能力。',
displayName: '360Zhinao2 o1',
enabled: true,
id: '360zhinao2-o1',
pricing: {
currency: 'CNY',
input: 4,
output: 10,
},
type: 'chat',
},
{
abilities: {
reasoning: true,
},
contextWindowTokens: 8000,
description:
'360gpt2-o1 使用树搜索构建思维链，并引入了反思机制，使用强化学习训练，模型具备自我反思与纠错的能力。',
@@ -10,8 +30,8 @@ const ai360ChatModels: AIChatModelCard[] = [
id: '360gpt2-o1',
pricing: {
currency: 'CNY',
input: 20,
output: 50,
input: 4,
output: 10,
},
type: 'chat',
},
src/config/aiModels/fireworksai.ts
+3
View File
@@ -235,6 +235,9 @@ const fireworksaiChatModels: AIChatModelCard[] = [
type: 'chat',
},
{
abilities: {
reasoning: true,
},
contextWindowTokens: 32_768,
description:
'Qwen QwQ 模型专注于推动 AI 推理，并展示了开放模型在推理能力上与闭源前沿模型匹敌的力量。QwQ-32B-Preview 是一个实验性发布版本，在 GPQA、AIME、MATH-500 和 LiveCodeBench 基准测试中，在分析和推理能力上可与 o1 相媲美，并超越 GPT-4o 和 Claude 3.5 Sonnet。注意：该模型目前作为无服务器模型进行实验性提供。如果用于生产环境，请注意 Fireworks 可能会在短时间内取消部署该模型。',
src/config/aiModels/giteeai.ts
+60
View File
@@ -1,6 +1,66 @@
import { AIChatModelCard } from '@/types/aiModel';
const giteeaiChatModels: AIChatModelCard[] = [
{
abilities: {
reasoning: true,
},
contextWindowTokens: 32_000,
description:
'基于 Qwen2.5-Math-1.5B 的 DeepSeek-R1 蒸馏模型，通过强化学习与冷启动数据优化推理性能，开源模型刷新多任务标杆。',
displayName: 'DeepSeek R1 Distill Qwen 1.5B',
enabled: true,
id: 'DeepSeek-R1-Distill-Qwen-1.5B',
type: 'chat',
},
{
abilities: {
reasoning: true,
},
contextWindowTokens: 32_000,
description:
'基于 Qwen2.5-Math-7B 的 DeepSeek-R1 蒸馏模型，通过强化学习与冷启动数据优化推理性能，开源模型刷新多任务标杆。',
displayName: 'DeepSeek R1 Distill Qwen 7B',
enabled: true,
id: 'DeepSeek-R1-Distill-Qwen-7B',
type: 'chat',
},
{
abilities: {
reasoning: true,
},
contextWindowTokens: 32_000,
description:
'基于 Qwen2.5-14B 的 DeepSeek-R1 蒸馏模型，通过强化学习与冷启动数据优化推理性能，开源模型刷新多任务标杆。',
displayName: 'DeepSeek R1 Distill Qwen 14B',
enabled: true,
id: 'DeepSeek-R1-Distill-Qwen-14B',
type: 'chat',
},
{
abilities: {
reasoning: true,
},
contextWindowTokens: 32_000,
description:
'DeepSeek-R1 系列通过强化学习与冷启动数据优化推理性能，开源模型刷新多任务标杆，超越 OpenAI-o1-mini 水平。',
displayName: 'DeepSeek R1 Distill Qwen 32B',
enabled: true,
id: 'DeepSeek-R1-Distill-Qwen-32B',
type: 'chat',
},
{
abilities: {
reasoning: true,
},
contextWindowTokens: 32_000,
description:
'QwQ-32B-Preview 是一款独具创新的自然语言处理模型，能够高效处理复杂的对话生成与上下文理解任务。',
displayName: 'QwQ 32B Preview',
enabled: true,
id: 'QwQ-32B-Preview',
type: 'chat',
},
{
abilities: {
functionCall: true,
src/config/aiModels/github.ts
+7
View File
@@ -4,6 +4,7 @@ const githubChatModels: AIChatModelCard[] = [
{
abilities: {
functionCall: true,
reasoning: true,
},
contextWindowTokens: 200_000,
description:
@@ -17,6 +18,7 @@ const githubChatModels: AIChatModelCard[] = [
},
{
abilities: {
reasoning: true,
vision: true,
},
contextWindowTokens: 128_000,
@@ -29,6 +31,7 @@ const githubChatModels: AIChatModelCard[] = [
},
{
abilities: {
reasoning: true,
vision: true,
},
contextWindowTokens: 200_000,
@@ -42,6 +45,7 @@ const githubChatModels: AIChatModelCard[] = [
},
{
abilities: {
reasoning: true,
vision: true,
},
contextWindowTokens: 128_000,
@@ -80,6 +84,9 @@ const githubChatModels: AIChatModelCard[] = [
type: 'chat',
},
{
abilities: {
reasoning: true,
},
contextWindowTokens: 128_000,
displayName: 'DeepSeek R1',
id: 'DeepSeek-R1',
src/config/aiModels/google.ts
+2
View File
@@ -4,6 +4,7 @@ const googleChatModels: AIChatModelCard[] = [
{
abilities: {
functionCall: true,
reasoning: true,
vision: true,
},
contextWindowTokens: 1_048_576 + 65_536,
@@ -24,6 +25,7 @@ const googleChatModels: AIChatModelCard[] = [
{
abilities: {
functionCall: true,
reasoning: true,
vision: true,
},
contextWindowTokens: 40_959,
src/config/aiModels/groq.ts
+12
View File
@@ -1,6 +1,18 @@
import { AIChatModelCard } from '@/types/aiModel';
const groqChatModels: AIChatModelCard[] = [
{
abilities: {
reasoning: true,
},
contextWindowTokens: 128_000,
description:
'DeepSeek R1——DeepSeek 套件中更大更智能的模型——被蒸馏到 Llama 70B 架构中。基于基准测试和人工评估，该模型比原始 Llama 70B 更智能，尤其在需要数学和事实精确性的任务上表现出色。',
displayName: 'DeepSeek R1 (Distil-Llama 70B)',
enabled: true,
id: 'deepseek-r1-distill-llama-70b',
type: 'chat',
},
{
abilities: {
functionCall: true,
src/config/aiModels/huggingface.ts
+6
View File
@@ -30,6 +30,9 @@ const huggingfaceChatModels: AIChatModelCard[] = [
type: 'chat',
},
{
abilities: {
reasoning: true,
},
contextWindowTokens: 32_768,
description: 'Qwen QwQ 是由 Qwen 团队开发的实验研究模型，专注于提升AI推理能力。',
displayName: 'QwQ 32B Preview',
@@ -50,6 +53,9 @@ const huggingfaceChatModels: AIChatModelCard[] = [
type: 'chat',
},
{
abilities: {
reasoning: true,
},
contextWindowTokens: 16_384,
displayName: 'DeepSeek R1',
enabled: true,
src/config/aiModels/internlm.ts
+19 -2
View File
@@ -7,7 +7,24 @@ const internlmChatModels: AIChatModelCard[] = [
},
contextWindowTokens: 32_768,
description:
'我们最新的模型系列，有着卓越的推理性能，支持 1M 的上下文长度以及更强的指令跟随和工具调用能力。',
'我们最新的模型系列，有着卓越的推理性能，领跑同量级开源模型。默认指向我们最新发布的 InternLM3 系列模型',
displayName: 'InternLM3',
enabled: true,
id: 'internlm3-latest',
maxOutput: 4096,
pricing: {
input: 0,
output: 0,
},
type: 'chat',
},
{
abilities: {
functionCall: true,
},
contextWindowTokens: 32_768,
description:
'我们仍在维护的老版本模型，经过多轮迭代有着极其优异且稳定的性能，包含 7B、20B 多种模型参数量可选，支持 1M 的上下文长度以及更强的指令跟随和工具调用能力。默认指向我们最新发布的 InternLM2.5 系列模型',
displayName: 'InternLM2.5',
enabled: true,
id: 'internlm2.5-latest',
@@ -23,7 +40,7 @@ const internlmChatModels: AIChatModelCard[] = [
functionCall: true,
},
contextWindowTokens: 32_768,
description: '我们仍在维护的老版本模型，有 7B、20B 多种模型参数量可选。',
description: 'InternLM2 版本最大的模型，专注于高度复杂的任务',
displayName: 'InternLM2 Pro Chat',
id: 'internlm2-pro-chat',
maxOutput: 4096,
src/config/aiModels/ollama.ts
+1
View File
@@ -75,6 +75,7 @@ const ollamaChatModels: AIChatModelCard[] = [
{
abilities: {
functionCall: true,
reasoning: true,
},
contextWindowTokens: 128_000,
description: 'QwQ 是一个实验研究模型，专注于提高 AI 推理能力。',
src/config/aiModels/openai.ts
+10
View File
@@ -11,6 +11,7 @@ export const openaiChatModels: AIChatModelCard[] = [
{
abilities: {
functionCall: true,
reasoning: true,
},
contextWindowTokens: 200_000,
description:
@@ -27,6 +28,9 @@ export const openaiChatModels: AIChatModelCard[] = [
type: 'chat',
},
{
abilities: {
reasoning: true,
},
contextWindowTokens: 128_000,
description:
'o1-mini是一款针对编程、数学和科学应用场景而设计的快速、经济高效的推理模型。该模型具有128K上下文和2023年10月的知识截止日期。',
@@ -42,6 +46,9 @@ export const openaiChatModels: AIChatModelCard[] = [
type: 'chat',
},
{
abilities: {
reasoning: true,
},
contextWindowTokens: 200_000,
description:
'o1是OpenAI新的推理模型，支持图文输入并输出文本，适用于需要广泛通用知识的复杂任务。该模型具有200K上下文和2023年10月的知识截止日期。',
@@ -57,6 +64,9 @@ export const openaiChatModels: AIChatModelCard[] = [
type: 'chat',
},
{
abilities: {
reasoning: true,
},
contextWindowTokens: 128_000,
description:
'o1是OpenAI新的推理模型，适用于需要广泛通用知识的复杂任务。该模型具有128K上下文和2023年10月的知识截止日期。',
src/config/aiModels/perplexity.ts
+3
View File
@@ -2,6 +2,9 @@ import { AIChatModelCard } from '@/types/aiModel';
const perplexityChatModels: AIChatModelCard[] = [
{
abilities: {
reasoning: true,
},
contextWindowTokens: 127_072,
description:
'由 DeepSeek 推理模型提供支持的新 API 产品。',
src/config/aiModels/qwen.ts
+2
View File
@@ -175,6 +175,7 @@ const qwenChatModels: AIChatModelCard[] = [
{
abilities: {
functionCall: true,
reasoning: 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。',
@@ -191,6 +192,7 @@ const qwenChatModels: AIChatModelCard[] = [
},
{
abilities: {
reasoning: true,
vision: true,
},
contextWindowTokens: 32_768,
src/config/aiModels/siliconcloud.ts
+4
View File
@@ -69,6 +69,7 @@ const siliconcloudChatModels: AIChatModelCard[] = [
},
{
abilities: {
reasoning: true,
vision: true,
},
contextWindowTokens: 32_768,
@@ -85,6 +86,9 @@ const siliconcloudChatModels: AIChatModelCard[] = [
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。',
src/config/aiModels/togetherai.ts
+64 -1
View File
@@ -1,6 +1,18 @@
import { AIChatModelCard } from '@/types/aiModel';
const togetheraiChatModels: AIChatModelCard[] = [
{
abilities: {
functionCall: true,
},
contextWindowTokens: 131_072,
description:
'Meta Llama 3.3 多语言大语言模型 ( LLM ) 是 70B（文本输入/文本输出）中的预训练和指令调整生成模型。 Llama 3.3 指令调整的纯文本模型针对多语言对话用例进行了优化，并且在常见行业基准上优于许多可用的开源和封闭式聊天模型。',
displayName: 'Llama 3.3 70B Instruct Turbo',
enabled: true,
id: 'meta-llama/Llama-3.3-70B-Instruct-Turbo',
type: 'chat',
},
{
contextWindowTokens: 131_072,
description:
@@ -245,15 +257,66 @@ const togetheraiChatModels: AIChatModelCard[] = [
id: 'microsoft/WizardLM-2-8x22B',
type: 'chat',
},
{
abilities: {
reasoning: true,
},
contextWindowTokens: 32_768,
description: 'DeepSeek-R1 系列通过强化学习与冷启动数据优化推理性能，开源模型刷新多任务标杆，超越 OpenAI-o1-mini 水平。',
displayName: 'DeepSeek-R1',
enabled: true,
id: 'deepseek-ai/DeepSeek-R1',
type: 'chat',
},
{
abilities: {
reasoning: true,
},
contextWindowTokens: 131_072,
description: 'DeepSeek-R1 蒸馏模型，通过强化学习与冷启动数据优化推理性能，开源模型刷新多任务标杆。',
displayName: 'DeepSeek R1 Distill Qwen 1.5B',
id: 'deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B',
type: 'chat',
},
{
abilities: {
reasoning: true,
},
contextWindowTokens: 131_072,
description: 'DeepSeek-R1 蒸馏模型，通过强化学习与冷启动数据优化推理性能，开源模型刷新多任务标杆。',
displayName: 'DeepSeek R1 Distill Qwen 14B',
id: 'deepseek-ai/DeepSeek-R1-Distill-Qwen-14B',
type: 'chat',
},
{
abilities: {
reasoning: true,
},
contextWindowTokens: 131_072,
description: 'DeepSeek-R1 蒸馏模型，通过强化学习与冷启动数据优化推理性能，开源模型刷新多任务标杆。',
displayName: 'DeepSeek R1 Distill Llama 70B',
id: 'deepseek-ai/DeepSeek-R1-Distill-Llama-70B',
type: 'chat',
},
{
contextWindowTokens: 16_384,
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',
enabled: true,
id: 'deepseek-ai/DeepSeek-V3',
type: 'chat',
},
{
contextWindowTokens: 4096,
description: 'DeepSeek LLM Chat (67B) 是创新的 AI 模型 提供深度语言理解和互动能力。',
displayName: 'DeepSeek LLM Chat (67B)',
enabled: true,
id: 'deepseek-ai/deepseek-llm-67b-chat',
type: 'chat',
},
{
abilities: {
reasoning: 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',
src/config/aiModels/zhipu.ts
+3
View File
@@ -2,6 +2,9 @@ import { AIChatModelCard } from '@/types/aiModel';
const zhipuChatModels: AIChatModelCard[] = [
{
abilities: {
reasoning: true,
},
contextWindowTokens: 16_384,
description: 'GLM-Zero-Preview具备强大的复杂推理能力，在逻辑推理、数学、编程等领域表现优异。',
displayName: 'GLM-Zero-Preview',