Text Generation Arena 文本生成模型排行榜

基于 Text Generation Arena 用户匿名投票的最新AI文本生成模型排行榜,涵盖各模型的 Elo 得分、95% 置信区间、投票量、机构与许可证。

榜首模型

Claude Fable 5

最高得分

1,508

模型数量

368

数据版本

2026年06月25日

数据来源: LM Arena

关于本排行榜

本排行榜展示了当前最强 AI 大模型在文本生成任务中的综合实力排名。数据来源于 LMArena(前身为 LMSYS Chatbot Arena),这是目前全球最大的 AI 模型众包评测平台。用户在平台上与两个匿名模型同时对话,并投票选出更好的回答——排名完全由真实用户的偏好决定,而非实验室基准测试。

评测方法概要

匿名盲测:用户同时与两个"隐藏身份"的模型对话,根据回答质量投票,排除品牌偏见。

Elo 评分:基于国际象棋领域的 Elo Rating 体系(Bradley-Terry 模型),通过对战结果计算每个模型的实力分数。分数越高,说明模型在真实对话中被用户选中的概率越大。

场景覆盖广泛:涵盖编程、创意写作、数学推理、知识问答、角色扮演等高频真实场景。

DataLearner 在原始数据基础上提供中文解读与深度分析,并将排行榜模型关联至 DataLearner 模型库,方便您一键查看模型详情、API 定价、评测得分等完整信息。

榜单历史快照月份:

排名总表

排名模型名称得分95% CI投票数机构许可证
AnthropicClaude Fable 5Anthropic1,508+/-94,366AnthropicProprietary
AnthropicClaude Opus 4.6 (thinking)Anthropic1,503+/-451,769AnthropicProprietary
AnthropicOpus 4.7 (thinking)Anthropic1,502+/-438,326AnthropicProprietary
4AnthropicClaude Opus 4.6Anthropic1,499+/-455,027AnthropicProprietary
5AnthropicOpus 4.7Anthropic1,494+/-439,550AnthropicProprietary
6Muse SparkFacebook AI研究实验室1,487+/-613,598Facebook AI研究实验室Proprietary
7Google Deep MindGemini 3.1 Pro PreviewGoogle Deep Mind1,486+/-468,291Google Deep MindProprietary
8Google Deep MindGemini 3.0 Pro (Preview 11-2025)Google Deep Mind1,486+/-441,298Google Deep MindProprietary
9AnthropicClaude Opus 4.8 (thinking)Anthropic1,484+/-618,680AnthropicProprietary
10OpenAIGPT-5.5 (high)OpenAI1,481+/-533,718OpenAIProprietary
11AnthropicClaude Opus 4.8Anthropic1,479+/-619,038AnthropicProprietary
12OpenAIGPT-5.4 (high)OpenAI1,478+/-446,702OpenAIProprietary
13Google Deep MindGemini 3.5 FlashGoogle Deep Mind1,476+/-710,159Google Deep MindProprietary
14OpenAIGPT-5.2OpenAI1,476+/-434,532OpenAIProprietary
15xAIgrok-4.20-beta-0309-reasoningxAI1,476+/-448,117xAIProprietary
16Qwen3.7-Max-Preview阿里巴巴1,475+/-103,731阿里巴巴Proprietary
17OpenAIGPT-5.5OpenAI1,475+/-534,794OpenAIProprietary
18xAIGrok 4.20 BetaxAI1,474+/-526,945xAIProprietary
19GLM 5.1智谱AI1,473+/-519,620智谱AIMIT
20Google Deep MindGemini 3.0 FlashGoogle Deep Mind1,473+/-430,704Google Deep MindProprietary
21AnthropicClaude Opus 4 (thinking-32k)Anthropic1,473+/-437,085AnthropicProprietary
22OpenAIGPT-5.5 InstantOpenAI1,473+/-526,215OpenAIProprietary
23AnthropicClaude Sonnet 4.6Anthropic1,472+/-445,162AnthropicProprietary
24xAIgrok-4.20-multi-agent-beta-0309xAI1,471+/-447,092xAIProprietary
25GLM-5.2 (max)智谱AI1,470+/-77,552智谱AIMIT
26AnthropicClaude Opus 4Anthropic1,469+/-371,133AnthropicProprietary
27ERNIE-5.1-Preview百度1,468+/-529,784百度Proprietary
28OpenAIGPT-5.4OpenAI1,467+/-449,383OpenAIProprietary
29mimo-v2.5-proXiaomi1,466+/-531,370XiaomiMIT
30xAIGrok 4.1 ThinkingxAI1,466+/-365,587xAIProprietary
31Qwen3.5 Max Preview阿里巴巴1,465+/-521,551阿里巴巴Proprietary
32Alibabaqwen3.7-plusAlibaba1,464+/-611,899AlibabaProprietary
33Moonshot AIKimi K2.6Moonshot AI1,461+/-529,798Moonshot AIModified MIT
34Qwen3.6-Max-Preview阿里巴巴1,460+/-85,217阿里巴巴Proprietary
35Google Deep MindGemini 3.0 Flash (minimal)Google Deep Mind1,460+/-372,062Google Deep MindProprietary
36xAIGrok 4.1xAI1,459+/-367,730xAIProprietary
37GLM-5智谱AI1,458+/-426,794智谱AIMIT
38DeepSeek-AIDeepSeek-V4-Pro (thinking)DeepSeek-AI1,457+/-531,874DeepSeek-AIMIT
39DeepSeek-AIDeepSeek-V4-ProDeepSeek-AI1,457+/-533,792DeepSeek-AIMIT
40AnthropicClaude Sonnet 4.5 (thinking-32k)Anthropic1,455+/-382,471AnthropicProprietary
41DOLA Seed 2.0 Pro字节跳动Seed团队1,455+/-455,915字节跳动Seed团队Proprietary
42AnthropicClaude Sonnet 4.5Anthropic1,455+/-380,906AnthropicProprietary
43OpenAIGPT-5.1 Pro (high)OpenAI1,455+/-440,817OpenAIProprietary
44DeepMindGemma 4 31BDeepMind1,451+/-85,892DeepMindApache 2.0
45Moonshot AIKimi K2 ThinkingMoonshot AI1,450+/-452,551Moonshot AIModified MIT
46AnthropicOpus 4.1 (thinking-16k)Anthropic1,449+/-349,794AnthropicProprietary
47ERNIE 5.0百度1,449+/-79,745百度Proprietary
48OpenAIGPT-5.4 mini (high)OpenAI1,449+/-445,301OpenAIProprietary
49OpenAIGPT-5.3OpenAI1,449+/-433,090OpenAIProprietary
50mimo-v2-proXiaomi1,449+/-524,588XiaomiProprietary
51MiniMaxminimax-m3MiniMax1,447+/-616,510MiniMaxProprietary
52ERNIE 5.0百度1,447+/-435,286百度Proprietary
53AnthropicOpus 4.1Anthropic1,447+/-377,297AnthropicProprietary
54Google Deep MindGemini 2.5 Pro Experimental 03-25Google Deep Mind1,446+/-3124,541Google Deep MindProprietary
55OpenAIGPT-4.5OpenAI1,445+/-614,547OpenAIProprietary
56Qwen 3.6 Plus Preview阿里巴巴1,444+/-533,663阿里巴巴Proprietary
57OpenAIGPT-4o(2025-03-27)OpenAI1,443+/-382,431OpenAIProprietary
58Qwen3.5-397B-A17B阿里巴巴1,443+/-447,774阿里巴巴Apache 2.0
59xAIGrok 4.3 BetaxAI1,443+/-533,981xAIProprietary
60GLM-4.7智谱AI1,442+/-612,114智谱AIMIT
61OpenAIGPT-5.1 InstantOpenAI1,439+/-443,443OpenAIProprietary
62DeepMindGemma 4 26B A4BDeepMind1,438+/-85,811DeepMindApache 2.0
63OpenAIGPT-5.2 Pro (high)OpenAI1,437+/-448,051OpenAIProprietary
64DeepSeek-AIDeepSeek-V4-Flash (thinking)DeepSeek-AI1,437+/-533,381DeepSeek-AIMIT
65Meituanlongcat-flash-chat-2602-expMeituan1,436+/-528,173MeituanProprietary
66DeepSeek-AIDeepSeek-V4-FlashDeepSeek-AI1,435+/-533,456DeepSeek-AIMIT
67OpenAIGPT-5.2OpenAI1,435+/-365,237OpenAIProprietary
68Qwen3 Max (Preview)阿里巴巴1,435+/-427,708阿里巴巴Proprietary
69OpenAIGPT-5-Pro (high)OpenAI1,434+/-531,926OpenAIProprietary
70mimo-v2.5Xiaomi1,434+/-532,112XiaomiMIT
71Moonshot AIKimi K2.5 InstantMoonshot AI1,432+/-78,179Moonshot AIModified MIT
72Googlegemini-3.1-flash-lite-previewGoogle1,432+/-454,156GoogleProprietary
73mimo-v2-omniXiaomi1,431+/-617,536XiaomiProprietary
74xAIGrok 4.1 Fast (fast-reasoning)xAI1,431+/-356,839xAIProprietary
75OpenAIOpenAI o3OpenAI1,431+/-459,742OpenAIProprietary
76Moonshot AIKimi K2 Thinking (thinking-turbo)Moonshot AI1,430+/-362,067Moonshot AIModified MIT
77Amazonamazon-nova-experimental-chat-26-02-10Amazon1,427+/-103,419AmazonProprietary
78Mistralmistral-medium-3.5Mistral1,427+/-710,777MistralModified MIT
79OpenAIGPT-5OpenAI1,427+/-431,562OpenAIProprietary
80GLM-4.6智谱AI1,425+/-435,626智谱AIMIT
81DeepSeek-AIDeepSeek V3.2DeepSeek-AI1,425+/-447,278DeepSeek-AIMIT
82DeepSeek-AIDeepSeek V3.2-Exp (thinking)DeepSeek-AI1,425+/-79,068DeepSeek-AIMIT
83AnthropicClaude Opus 4 (thinking-16k)Anthropic1,424+/-436,886AnthropicProprietary
84Alibabaqwen3-max-2025-09-23Alibaba1,424+/-69,155AlibabaProprietary
85Qwen3-235B-A22B-2507阿里巴巴1,423+/-397,193阿里巴巴Apache 2.0
86DeepSeek-AIDeepSeek V3.2 (thinking)DeepSeek-AI1,423+/-441,077DeepSeek-AIMIT
87DeepSeek-AIDeepSeek V3.2-ExpDeepSeek-AI1,423+/-611,923DeepSeek-AIMIT
88DeepSeek-AIDeepSeek-R1-0528DeepSeek-AI1,422+/-618,463DeepSeek-AIMIT
89xAIGrok 4 FastxAI1,421+/-86,810xAIProprietary
90Nvidianvidia-nemotron-3-ultra-550b-a55b-nvfp4Nvidia1,420+/-77,484NvidiaOpenMDW-1.1
91ERNIE 5.0百度1,419+/-94,704百度Proprietary
92Moonshot AIKimi K2 0905Moonshot AI1,418+/-611,771Moonshot AIModified MIT
93DeepSeek-AIDeepSeek-V3.1 Terminus (thinking)DeepSeek-AI1,418+/-103,461DeepSeek-AIMIT
94Moonshot AIKimi K2Moonshot AI1,417+/-527,629Moonshot AIModified MIT
95DeepSeek-AIDeepSeek-V3.1DeepSeek-AI1,417+/-614,957DeepSeek-AIMIT
96Qwen3.5-122B-A10B阿里巴巴1,417+/-428,566阿里巴巴Apache 2.0
97DeepSeek-AIDeepSeek-V3.1 (thinking)DeepSeek-AI1,417+/-711,730DeepSeek-AIMIT
98MiniMaxAIMiniMax-M2.7MiniMaxAI1,417+/-439,341MiniMaxAIModified MIT
99DeepSeek-AIDeepSeek-V3.1 TerminusDeepSeek-AI1,416+/-103,700DeepSeek-AIMIT
100Amazonamazon-nova-experimental-chat-26-01-10Amazon1,416+/-103,407AmazonProprietary
101MistralAIMistral Large 3MistralAI1,416+/-444,098MistralAIApache 2.0
102Qwen3-VL-235B-A22B-Instruct阿里巴巴1,415+/-611,508阿里巴巴Apache 2.0
103OpenAIGPT-4.1OpenAI1,414+/-450,973OpenAIProprietary
104Tencenthunyuan-hy3-previewTencent1,413+/-86,666Tencenttencent-hunyuan-community
105AnthropicClaude Opus 4Anthropic1,412+/-444,197AnthropicProprietary
106xAIGrok 3xAI1,412+/-432,909xAIProprietary
107AnthropicHaiku 4.5Anthropic1,411+/-396,743AnthropicProprietary
108GLM-4.5智谱AI1,411+/-524,302智谱AIMIT
109Google Deep MindGemini 2.5 FlashGoogle Deep Mind1,410+/-2124,489Google Deep MindProprietary
110xAIGrok 4xAI1,410+/-441,363xAIProprietary
111MistralAIMagistral-Medium-2506MistralAI1,410+/-393,957MistralAIProprietary
112Qwen3.5-27B阿里巴巴1,409+/-427,413阿里巴巴Apache 2.0
113Google Deep MindGemini 2.5 Flash-Preview-09-2025Google Deep Mind1,404+/-432,892Google Deep MindProprietary
114xAIGrok 4 Fast (fast-reasoning)xAI1,404+/-518,707xAIProprietary
115OpenAIGPT-5.4 nano (high)OpenAI1,403+/-444,349OpenAIProprietary
116Alibabaqwen3-235b-a22b-no-thinkingAlibaba1,403+/-538,199AlibabaApache 2.0
117OpenAIOpenAI o1OpenAI1,402+/-427,807OpenAIProprietary
118Qwen3-Next阿里巴巴1,401+/-522,875阿里巴巴Apache 2.0
119Meituanlongcat-flash-chatMeituan1,401+/-611,392MeituanMIT
120Alibabaqwen3-235b-a22b-thinking-2507Alibaba1,399+/-78,991AlibabaApache 2.0
121AnthropicClaude Sonnet 4 (thinking-32k)Anthropic1,399+/-435,101AnthropicProprietary
122DeepSeek-AIDeepSeek-R1DeepSeek-AI1,398+/-518,524DeepSeek-AIMIT
123StepFunAIStep 3.5 FlashStepFunAI1,397+/-445,836StepFunAIProprietary
124Qwen3.5-35B-A3B阿里巴巴1,396+/-429,232阿里巴巴Apache 2.0
125Qwen3-VL-235B-A22B-Instruct (thinking)阿里巴巴1,396+/-77,942阿里巴巴Apache 2.0
126Tencenthunyuan-vision-1.5-thinkingTencent1,396+/-122,217TencentProprietary
127DeepSeek-AIDeepSeek-V3-0324DeepSeek-AI1,395+/-445,500DeepSeek-AIMIT
128StepFunAIStep 3.5 FlashStepFunAI1,395+/-449,407StepFunAIApache 2.0
129Amazonamazon-nova-experimental-chat-12-10Amazon1,395+/-103,675AmazonProprietary
130mimo-v2-flash (non-thinking)Xiaomi1,393+/-446,656XiaomiMIT
131MiniMaxAIMiniMax M2.5MiniMaxAI1,391+/-441,235MiniMaxAIModified MIT
132OpenAIGPT-5-mini (high)OpenAI1,390+/-527,022OpenAIProprietary
133OpenAIOpenAI o4 - miniOpenAI1,390+/-445,437OpenAIProprietary
134AnthropicClaude Sonnet 4Anthropic1,389+/-440,301AnthropicProprietary
135OpenAIOpenAI o1OpenAI1,388+/-531,122OpenAIProprietary
136Qwen3-Coder-480B-A35B阿里巴巴1,388+/-525,714阿里巴巴Apache 2.0
137mimo-v2-flash (thinking)Xiaomi1,387+/-610,956XiaomiMIT
138AnthropicClaude Sonnet 3.7 (thinking-32k)Anthropic1,387+/-438,820AnthropicProprietary
139Hunyuan-T1腾讯AI实验室1,387+/-94,704腾讯AI实验室Proprietary
140Mistralmistral-medium-2505Mistral1,387+/-533,219MistralProprietary
141MiniMaxAIM2.1MiniMaxAI1,385+/-517,113MiniMaxAIMIT
142Qwen3-30B-A3B-2507阿里巴巴1,383+/-523,720阿里巴巴Apache 2.0
143OpenAIGPT-4.1 miniOpenAI1,383+/-439,317OpenAIProprietary
144Tencenthunyuan-turbos-20250416Tencent1,382+/-610,724TencentProprietary
145Google Deep MindGemini 2.5 Flash-Lite-Preview-09-2025 (no-thinking)Google Deep Mind1,380+/-347,216Google Deep MindProprietary
146trinity-large-previewArcee AI1,379+/-430,121Arcee AIApache 2.0
147GLM-4.6V智谱AI1,377+/-112,806智谱AIMIT
148Qwen3-235B-A22B阿里巴巴1,375+/-526,264阿里巴巴Apache 2.0
149Google Deep MindGemini 2.5 Flash-Lite (thinking)Google Deep Mind1,374+/-532,886Google Deep MindProprietary
150Qwen2.5-Max阿里巴巴1,374+/-432,619阿里巴巴Proprietary
151GLM-4.5-Air智谱AI1,373+/-431,076智谱AIMIT
152AnthropicClaude 3.5 SonnetAnthropic1,373+/-388,336AnthropicProprietary
153AnthropicClaude Sonnet 3.7Anthropic1,371+/-443,179AnthropicProprietary
154Qwen3-Next (thinking)阿里巴巴1,370+/-613,693阿里巴巴Apache 2.0
155trinity-large-thinkingArcee AI1,369+/-529,269Arcee AIApache 2.0
156GLM-4.7-Flash智谱AI1,368+/-611,722智谱AIMIT
157Amazonamazon-nova-experimental-chat-11-10Amazon1,367+/-425,375AmazonProprietary
158Google Deep MindGemma 3 - 27B (IT)Google Deep Mind1,366+/-447,523Google Deep MindGemma
159MiniMaxminimax-m1MiniMax1,364+/-435,184MiniMaxApache 2.0
160OpenAIOpenAI o3-mini (high)OpenAI1,363+/-518,589OpenAIProprietary
161OpenAIOpenAI o3-mini (high)OpenAI1,362+/-516,963OpenAIProprietary
162Nvidianvidia-nemotron-3-super-120b-a12bNvidia1,362+/-77,547NvidiaNVIDIA Open Model
163DeepMindGemini 2.0 Flash ExperimentalDeepMind1,360+/-443,747DeepMindProprietary
164DeepSeek-AIDeepSeek-V3DeepSeek-AI1,358+/-521,770DeepSeek-AIDeepSeek
165MistralAIMistral-Small-3.2MistralAI1,358+/-517,705MistralAIApache 2.0
166xAIgrok-3-mini-betaxAI1,357+/-522,714xAIProprietary
167intellect-3Prime Intellect1,356+/-85,330Prime IntellectMIT
168CohereAIC4AI Command A (202503)CohereAI1,354+/-356,246CohereAICC-BY-NC-4.0
169DeepMindGemini 2.0 Flash-LiteDeepMind1,353+/-424,955DeepMindProprietary
170GLM-4.5V智谱AI1,353+/-84,959智谱AIMIT
171OpenAIGPT OSS 120BOpenAI1,353+/-430,629OpenAIApache 2.0
172Google Deep MindGemini 1.5 ProGoogle Deep Mind1,351+/-355,606Google Deep MindProprietary
173Amazonamazon-nova-experimental-chat-10-20Amazon1,350+/-611,467AmazonProprietary
174Tencenthunyuan-turbos-20250226Tencent1,349+/-122,220TencentProprietary
175StepFunAIStep3StepFunAI1,348+/-76,542StepFunAIApache 2.0
176OpenAIOpenAI o3-miniOpenAI1,348+/-457,336OpenAIProprietary
177Amazonamazon-nova-experimental-chat-10-09Amazon1,348+/-112,837AmazonProprietary
178Nvidiallama-3.1-nemotron-ultra-253b-v1Nvidia1,347+/-122,549NvidiaNvidia Open Model
179Qwen3-32B阿里巴巴1,347+/-93,926阿里巴巴Apache 2.0
180mercury-2Inception AI1,347+/-113,124Inception AIProprietary
181ling-flash-2.0InclusionAI1,346+/-76,999InclusionAIMIT
182Alibabaqwen-plus-0125Alibaba1,346+/-85,819AlibabaProprietary
183MiniMaxAIMiniMax M2MiniMaxAI1,346+/-86,866MiniMaxAIApache 2.0
184OpenAIGPT-4oOpenAI1,346+/-3112,881OpenAIProprietary
185Nvidianvidia-llama-3.3-nemotron-super-49b-v1.5Nvidia1,343+/-103,346NvidiaNvidia Open
186glm-4-plus-0111Zhipu1,343+/-85,760ZhipuProprietary
187AnthropicClaude 3.5 SonnetAnthropic1,342+/-382,419AnthropicProprietary
188Google Deep MindGemma 3 - 12B (IT)Google Deep Mind1,342+/-103,829Google Deep MindGemma
189Tencenthunyuan-turbo-0110Tencent1,341+/-122,290TencentProprietary
190OpenAIGPT-5-Nano (high)OpenAI1,337+/-78,266OpenAIProprietary
191OpenAIOpenAI o1-miniOpenAI1,337+/-451,981OpenAIProprietary
192Nova 2 Lite亚马逊1,337+/-612,226亚马逊Proprietary
193QwQ-32B阿里巴巴1,336+/-425,392阿里巴巴Apache 2.0
194xAIGrok 2xAI1,336+/-463,498xAIProprietary
195Googlegemini-advanced-0514Google1,335+/-550,148GoogleProprietary
196OpenAIGPT-4oOpenAI1,335+/-445,499OpenAIProprietary
197Metallama-3.1-405b-instruct-bf16Meta1,335+/-441,375MetaLlama 3.1 Community
198StepFunstep-2-16k-exp-202412StepFun1,334+/-94,833StepFunProprietary
199Metallama-3.1-405b-instruct-fp8Meta1,333+/-459,656MetaLlama 3.1 Community
200olmo-3.1-32b-instructAi21,330+/-612,217Ai2Apache 2.0
201molmo-2-8bAi21,328+/-21803Ai2Apache 2.0
202yi-lightning01 AI1,328+/-527,33201 AIProprietary
203Nvidiallama-3.3-nemotron-49b-super-v1Nvidia1,328+/-122,218NvidiaNvidia
204Qwen3-30B-A3B阿里巴巴1,327+/-526,486阿里巴巴Apache 2.0
205Llama 4 Maverick InstructFacebook AI研究实验室1,327+/-439,977Facebook AI研究实验室Llama 4
206Tencenthunyuan-large-2025-02-10Tencent1,326+/-103,738TencentProprietary
207OpenAIgpt-4-turbo-2024-04-09OpenAI1,324+/-498,114OpenAIProprietary
208AnthropicClaude 3.5 HaikuAnthropic1,324+/-369,967AnthropicProprietary
209Google Deep MindGemini 1.5 ProGoogle Deep Mind1,323+/-479,138Google Deep MindProprietary
210DeepSeekdeepseek-v2.5-1210DeepSeek1,323+/-86,795DeepSeekDeepSeek
211Llama 4 Scout InstructFacebook AI研究实验室1,323+/-530,288Facebook AI研究实验室Llama
212OpenAIGPT-4.1 nanoOpenAI1,322+/-86,103OpenAIProprietary
213AnthropicClaude3-OpusAnthropic1,321+/-3194,909AnthropicProprietary
214ring-flash-2.0InclusionAI1,321+/-77,144InclusionAIMIT
215StepFunstep-1o-turbo-202506StepFun1,320+/-79,040StepFunProprietary
216glm-4-plusZhipu AI1,319+/-526,126Zhipu AIProprietary
217Llama3.3-70B-InstructFacebook AI研究实验室1,318+/-354,732Facebook AI研究实验室Llama-3.3
218Google Deep MindGemma-3n-E4BGoogle Deep Mind1,318+/-522,583Google Deep MindGemma
219Alibabaqwen-max-0919Alibaba1,318+/-616,478AlibabaQwen
220OpenAIGPT-4o miniOpenAI1,318+/-468,709OpenAIProprietary
221OpenAIGPT OSS 20BOpenAI1,317+/-610,625OpenAIApache 2.0
222Nvidianvidia-nemotron-3-nano-30b-a3b-bf16Nvidia1,316+/-615,510NvidiaNVIDIA Open Model
223Alibabaqwen2.5-plus-1127Alibaba1,315+/-610,187AlibabaProprietary
224athene-v2-chatNexusFlow1,314+/-524,739NexusFlowNexusFlow
225Mistralmistral-large-2407Mistral1,314+/-445,459MistralMistral Research
226OpenAIGPT-4OpenAI1,313+/-493,439OpenAIProprietary
227OpenAIGPT-4OpenAI1,312+/-4100,105OpenAIProprietary
228Tencenthunyuan-standard-2025-02-10Tencent1,311+/-103,904TencentProprietary
229Googlegemini-1.5-flash-002Google1,309+/-434,902GoogleProprietary
230xAIgrok-2-mini-2024-08-13xAI1,308+/-452,567xAIProprietary
231granite-4.1-8bIBM1,307+/-104,072IBMApache 2.0
232DeepSeek-AIDeepSeek V2.5DeepSeek-AI1,307+/-524,572DeepSeek-AIDeepSeek
233mercuryInception AI1,306+/-141,953Inception AIProprietary
234athene-70b-0725NexusFlow1,306+/-619,621NexusFlowCC-BY-NC-4.0
235olmo-3-32b-thinkAi21,306+/-85,944Ai2Apache 2.0
236Mistralmistral-large-2411Mistral1,305+/-428,073MistralMRL
237MistralAIMagistral-Medium-2506MistralAI1,304+/-611,635MistralAIProprietary
238MistralAIMistral-Small-3.1-24B-Instruct-2503MistralAI1,303+/-533,210MistralAIApache 2.0
239Google Deep MindGemma 3 - 4B (IT)Google Deep Mind1,303+/-94,171Google Deep MindGemma
240Qwen2.5-VL-72B-Instruct阿里巴巴1,303+/-439,406阿里巴巴Qwen
241Llama3.1-70B-InstructFacebook AI研究实验室1,299+/-87,140Facebook AI研究实验室Llama 3.1
242Tencenthunyuan-large-visionTencent1,294+/-95,371TencentProprietary
243Llama3.1-70B-InstructFacebook AI研究实验室1,293+/-455,240Facebook AI研究实验室Llama 3.1 Community
244Amazonamazon-nova-pro-v1.0Amazon1,290+/-524,745AmazonProprietary
245jamba-1.5-largeAI21 Labs1,289+/-78,662AI21 LabsJamba Open
246Googlegemma-2-27b-itGoogle1,289+/-375,754GoogleGemma license
247reka-core-20240904Reka AI1,288+/-77,312Reka AIProprietary
248ibm-granite-h-smallIBM1,287+/-85,683IBMApache 2.0
249OpenAIGPT-4OpenAI1,287+/-554,173OpenAIProprietary
250Googlegemini-1.5-flash-001Google1,286+/-562,833GoogleProprietary
251llama-3.1-tulu-3-70bAi21,286+/-102,846Ai2Llama 3.1
252Nvidiallama-3.1-nemotron-51b-instructNvidia1,286+/-103,749NvidiaLlama 3.1
253olmo-3.1-32b-thinkAi21,285+/-78,499Ai2Apache 2.0
254AnthropicClaude3-SonnetAnthropic1,280+/-4109,284AnthropicProprietary
255gemma-2-9b-it-simpoPrinceton1,280+/-710,072PrincetonMIT
256Nvidianemotron-4-340b-instructNvidia1,276+/-519,659NvidiaNVIDIA Open Model
257Llama3-70B-InstructFacebook AI研究实验室1,276+/-4156,876Facebook AI研究实验室Llama 3 Community
258Coherecommand-r-plus-08-2024Cohere1,276+/-79,866CohereCC-BY-NC-4.0
259OpenAIGPT-4OpenAI1,275+/-488,723OpenAIProprietary
260MistralAIMistral Small 24B Instruct 2501MistralAI1,274+/-614,681MistralAIApache 2.0
261GLM4智谱AI1,273+/-79,788智谱AIProprietary
262reka-flash-20240904Reka AI1,272+/-77,536Reka AIProprietary
263Qwen2.5-Coder-32B-Instruct阿里巴巴1,270+/-85,432阿里巴巴Apache 2.0
264CohereAIC4AI Aya Vision 32BCohereAI1,267+/-527,124CohereAICC-BY-NC-4.0
265Googlegemma-2-9b-itGoogle1,266+/-454,611GoogleGemma license
266DeepSeekdeepseek-coder-v2DeepSeek1,264+/-615,147DeepSeekDeepSeek License
267Qwen2-72B-Instruct阿里巴巴1,261+/-537,325阿里巴巴Qianwen LICENSE
268CohereAIC4AI Command R+CohereAI1,261+/-477,554CohereAICC-BY-NC-4.0
269AnthropicClaude3-HaikuAnthropic1,261+/-4117,701AnthropicProprietary
270Amazonamazon-nova-lite-v1.0Amazon1,260+/-519,372AmazonProprietary
271Googlegemini-1.5-flash-8b-001Google1,259+/-435,558GoogleProprietary
272Microsoft AzurePhi 4 - 14BMicrosoft Azure1,256+/-524,126Microsoft AzureMIT
273olmo-2-0325-32b-instructAi21,251+/-113,334Ai2Apache-2.0
274Coherecommand-r-08-2024Cohere1,250+/-710,140CohereCC-BY-NC-4.0
275Mistralmistral-large-2402Mistral1,242+/-562,436MistralProprietary
276Amazonamazon-nova-micro-v1.0Amazon1,241+/-519,364AmazonProprietary
277jamba-1.5-miniAI21 Labs1,239+/-78,858AI21 LabsJamba Open
278Mistralministral-8b-2410Mistral1,237+/-94,781MistralMRL
279Googlegemini-pro-dev-apiGoogle1,235+/-718,354GoogleProprietary
280Qwen1.5-110B-Chat阿里巴巴1,233+/-626,195阿里巴巴Qianwen LICENSE
281Tencenthunyuan-standard-256kTencent1,233+/-122,728TencentProprietary
282reka-flash-21b-20240226-onlineReka AI1,233+/-715,450Reka AIProprietary
283Qwen1.5-72B-Chat阿里巴巴1,233+/-539,302阿里巴巴Qianwen LICENSE
284MistralAIMixtral-8x22B-Instruct-v0.1MistralAI1,229+/-551,416MistralAIApache 2.0
285Coherecommand-rCohere1,226+/-554,036CohereCC-BY-NC-4.0
286reka-flash-21b-20240226Reka AI1,226+/-624,806Reka AIProprietary
287OpenAIgpt-3.5-turbo-0125OpenAI1,224+/-566,207OpenAIProprietary
288Llama3-8B-InstructFacebook AI研究实验室1,223+/-4104,642Facebook AI研究实验室Llama 3 Community
289CohereAIC4AI Aya Vision 8BCohereAI1,223+/-79,818CohereAICC-BY-NC-4.0
290DeepMindGemini-proDeepMind1,222+/-126,390DeepMindProprietary
291Mistralmistral-mediumMistral1,222+/-634,550MistralProprietary
292llama-3.1-tulu-3-8bAi21,220+/-112,896Ai2Llama 3.1
293Yi-1.5-34B零一万物1,212+/-524,146零一万物Apache-2.0
294zephyr-orpo-141b-A35b-v0.1HuggingFace1,212+/-114,652HuggingFaceApache 2.0
295Llama3.1-8B-InstructFacebook AI研究实验室1,211+/-449,605Facebook AI研究实验室Llama 3.1 Community
296Llama3.1-8B-InstructFacebook AI研究实验室1,208+/-113,090Facebook AI研究实验室Apache 2.0
297Alibabaqwen1.5-32b-chatAlibaba1,203+/-621,741AlibabaQianwen LICENSE
298OpenAIgpt-3.5-turbo-1106OpenAI1,202+/-916,619OpenAIProprietary
299Googlegemma-2-2b-itGoogle1,200+/-446,616GoogleGemma license
300Microsoft AzurePhi-3-medium 14B-previewMicrosoft Azure1,197+/-525,055Microsoft AzureMIT
301Mistralmixtral-8x7b-instruct-v0.1Mistral1,196+/-473,503MistralApache 2.0
302DBRX Instructdatabricks1,194+/-632,191databricksDBRX LICENSE
303InternLM2-Base-20B上海人工智能实验室1,191+/-79,901上海人工智能实验室Other
304Qwen1.5-14B-Chat阿里巴巴1,190+/-717,839阿里巴巴Qianwen LICENSE
305WizardLM-70B-V1.0WizardLM Team1,184+/-98,214WizardLM TeamLlama 2 Community
306DeepSeek-AIDeepSeek LLM 67B ChatDeepSeek-AI1,184+/-124,932DeepSeek-AIDeepSeek License
307Yi-34B零一万物1,183+/-715,483零一万物Yi License
308granite-3.0-8b-instructIBM1,182+/-96,638IBMApache 2.0
309openchat-3.5OpenChat1,182+/-107,968OpenChatApache-2.0
310openchat-3.5-0106OpenChat1,182+/-812,637OpenChatApache-2.0
311Google ResearchGemma 1.1-7B-ITGoogle Research1,181+/-623,893Google ResearchGemma license
312snowflake-arctic-instructSnowflake1,179+/-632,832SnowflakeApache 2.0
313granite-3.1-2b-instructIBM1,178+/-113,188IBMApache 2.0
314tulu-2-dpo-70bAllenAI/UW1,177+/-106,535AllenAI/UWAI2 ImpACT Low-risk
315openhermes-2.5-mistral-7bNousResearch1,175+/-105,006NousResearchApache-2.0
316Vicuna 33BLM-SYS1,172+/-622,479LM-SYSNon-commercial
317starling-lm-7b-betaNexusflow1,171+/-716,056NexusflowApache-2.0
318Microsoft AzurePhi-3-small 7BMicrosoft Azure1,170+/-617,766Microsoft AzureMIT
319Metallama-2-70b-chatMeta1,170+/-638,492MetaLlama 2 Community
320starling-lm-7b-alphaUC Berkeley1,167+/-810,224UC BerkeleyCC-BY-NC-4.0
321Metallama-3.2-3b-instructMeta1,166+/-87,936MetaLlama 3.2
322nous-hermes-2-mixtral-8x7b-dpoNousResearch1,164+/-123,777NousResearchApache-2.0
323Qwen3-VL-2B阿里巴巴1,156+/-86,837阿里巴巴Apache 2.0
324QwQ-32B-Preview阿里巴巴1,155+/-113,231阿里巴巴Apache 2.0
325Nvidiallama2-70b-steerlm-chatNvidia1,154+/-133,585NvidiaLlama 2 Community
326solar-10.7b-instruct-v1.0Upstage AI1,151+/-134,155Upstage AICC-BY-NC-4.0
327dolphin-2.2.1-mistral-7bCognitive Computations1,151+/-151,679Cognitive ComputationsApache-2.0
328MPT-30B-ChatMosaicML1,150+/-122,572MosaicMLCC-BY-NC-SA-4.0
329MistralAIMistral-7B-Instruct-v0.2MistralAI1,149+/-719,402MistralAIApache-2.0
330Microsoftwizardlm-13bMicrosoft1,148+/-97,044MicrosoftLlama 2 Community
331falcon-180b-chatTII1,147+/-171,295TIIFalcon-180B TII License
332Qwen1.5-7B-Chat阿里巴巴1,143+/-104,737阿里巴巴Qianwen LICENSE
333Microsoft AzurePhi-3-mini 3.8BMicrosoft Azure1,142+/-612,297Microsoft AzureMIT
334Baichuan2-13B-Chat百川智能1,141+/-719,174百川智能Llama 2 Community
335Vicuna 13BLM-SYS1,140+/-719,367LM-SYSLlama 2 Community
336Qwen-14B-Chat阿里巴巴1,138+/-114,964阿里巴巴Qianwen LICENSE
337Google ResearchPaLM 2Google Research1,137+/-98,554Google ResearchProprietary
338Google ResearchGemma 7B - ItGoogle Research1,137+/-98,925Google ResearchGemma license
339CodeLLaMA-34BFacebook AI研究实验室1,136+/-97,366Facebook AI研究实验室Llama 2 Community
340zephyr-7b-betaHuggingFace1,130+/-911,118HuggingFaceMIT
341Microsoft AzurePhi-3-mini 3.8BMicrosoft Azure1,129+/-720,685Microsoft AzureMIT
342Microsoft AzurePhi-3-mini 3.8BMicrosoft Azure1,127+/-620,118Microsoft AzureMIT
343guanaco-33bUW1,126+/-122,921UWNon-commercial
344zephyr-7b-alphaHuggingFace1,126+/-161,785HuggingFaceMIT
345stripedhyena-nous-7bTogether AI1,120+/-115,182Together AIApache 2.0
346CodeLlama-70B-InstructFacebook AI研究实验室1,118+/-181,143Facebook AI研究实验室Llama 2 Community
347Google ResearchGemma 1.1-2B-ITGoogle Research1,115+/-810,854Google ResearchGemma license
348Vicuna 7BLM-SYS1,114+/-96,923LM-SYSLlama 2 Community
349smollm2-1.7b-instructHuggingFace1,114+/-142,199HuggingFaceApache 2.0
350Metallama-3.2-1b-instructMeta1,110+/-88,045MetaLlama 3.2
351MistralAIMistral 7B InstructMistralAI1,109+/-98,977MistralAIApache 2.0
352Baichuan2-7B-Chat百川智能1,107+/-714,148百川智能Llama 2 Community
353Google ResearchGemma 2B - ItGoogle Research1,092+/-114,780Google ResearchGemma license
354Qwen1.5-4B-Chat阿里巴巴1,090+/-97,597阿里巴巴Qianwen LICENSE
355olmo-7b-instructAi21,073+/-116,328Ai2Apache-2.0
356Koala达摩院1,070+/-106,965达摩院Non-commercial
357alpaca-13bStanford1,068+/-115,745StanfordNon-commercial
358GPT4All 13BNomic AI1,066+/-151,743Nomic AINon-commercial
359MPT-7B-ChatMosaicML1,062+/-123,924MosaicMLCC-BY-NC-SA-4.0
360ChatGLM3-6B智谱AI1,055+/-124,658智谱AIApache-2.0
361RWKV-4-Raven-14BRWKV1,041+/-114,845RWKVApache 2.0
362ChatGLM2-6B智谱AI1,023+/-142,658智谱AIApache-2.0
363oasst-pythia-12bOpenAssistant1,022+/-116,310OpenAssistantApache 2.0
364ChatGLM-6B智谱AI995+/-134,914智谱AINon-commercial
365fastchat-t5-3bLMSYS991+/-124,203LMSYSApache 2.0
366dolly-v2-12bDatabricks980+/-143,412DatabricksMIT
367LLaMA 13BFacebook AI研究实验室973+/-162,391Facebook AI研究实验室Non-commercial
368stablelm-tuned-alpha-7bStability AI952+/-133,287Stability AICC-BY-NC-SA-4.0

数据仅供参考,以官方来源为准。模型名称旁的链接可跳转到 DataLearner 模型详情页。

常见问题 (FAQ)

01

什么是 Text Generation Arena (LMArena)?

Text Generation Arena(原 LMSYS Chatbot Arena)是目前最具影响力的大模型匿名评测平台。用户向两个身份未知的模型提问,根据回答质量投票,系统通过 Elo 算法将数百万次投票汇聚为动态排行榜,被学术界和工业界广泛引用。

02

Arena Elo 分数是如何计算的?

Elo 算法源自国际象棋评分体系。每次对战后,胜者得分上升、败者下降,幅度取决于双方原始评分差距。95% 置信区间(CI)反映该模型参与对战次数的多少:CI 越窄说明数据越充分、排名越可信。

03

为什么同一模型会出现"Thinking"和普通两个版本?

部分模型支持"扩展思考"(Extended Thinking)模式,会在给出最终答案前进行更深入的内部推理。该模式通常在逻辑推理、数学和编程任务上得分更高,但响应时延也更长、成本更高。Arena 将两种模式分开评测,以便用户根据实际需求选择。

04

如何根据排行榜选择适合自己的大语言模型?

建议综合考虑:综合性能(看 Elo 总分)、成本(闭源 API 按量计费,开源可自部署)、中文支持开源程度以及响应速度