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
Gemini 4 Argon (High)
最高得分
1,524
模型数量
416
数据版本
2026年10月15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。
排名总表
| 排名 | 模型名称 | 得分 | 95% CI | 投票数 | 机构 | 许可证 |
|---|---|---|---|---|---|---|
 Gemini 4 Argon (High)Google | 1,524 | +/-9 | 4,904 | Proprietary | ||
 Muse Wonder (Max)Meta | 1,509 | +/-11 | 3,247 | Meta | Proprietary | |
 Claude Opus 5.5 (High)Anthropic | 1,508 | +/-7 | 8,300 | Anthropic | Proprietary | |
| 4 | Claude Fable 5 (High)Anthropic | 1,504 | +/-4 | 42,004 | Anthropic | Proprietary |
| 5 | Claude Opus 4.6 (high)Anthropic | 1,503 | +/-3 | 81,394 | Anthropic | Proprietary |
| 6 | Claude Fable 5.1 (Max)Anthropic | 1,502 | +/-6 | 14,590 | Anthropic | Proprietary |
| 7 | Claude Opus 4.7 (High)Anthropic | 1,501 | +/-4 | 65,116 | Anthropic | Proprietary |
| 8 | Claude Opus 4.6Anthropic | 1,497 | +/-3 | 86,071 | Anthropic | Proprietary |
| 9 | Gemini 3.8 Flash (High)Google | 1,496 | +/-5 | 37,454 | Proprietary | |
| 10 | Muse Spark 1.3 (Max)Meta | 1,494 | +/-6 | 17,561 | Meta | Proprietary |
| 11 | Opus 4.7Anthropic | 1,494 | +/-4 | 66,216 | Anthropic | Proprietary |
| 12 | Muse Spark 1.2 (xhigh)Facebook AI研究实验室 | 1,493 | +/-7 | 8,248 | Facebook AI研究实验室 | Proprietary |
| 13 | Muse Spark 1.1Facebook AI研究实验室 | 1,490 | +/-4 | 42,810 | Facebook AI研究实验室 | Proprietary |
| 14 | Claude Opus 5 (max)Anthropic | 1,490 | +/-5 | 32,495 | Anthropic | Proprietary |
| 15 | Claude Opus 5 (high)Anthropic | 1,490 | +/-4 | 66,824 | Anthropic | Proprietary |
| 16 | Muse SparkFacebook AI研究实验室 | 1,489 | +/-6 | 14,150 | Facebook AI研究实验室 | Proprietary |
| 17 | Gemini 3.7 Flash (high)Google DeepMind | 1,488 | +/-5 | 27,232 | Google DeepMind | Proprietary |
| 18 |  Kimi K3 (max)Moonshot AI | 1,487 | +/-5 | 32,795 | Moonshot AI | Kimi K3 license |
| 19 | Gemini 3.1 Pro PreviewGoogle DeepMind | 1,487 | +/-3 | 127,649 | Google DeepMind | Proprietary |
| 20 | Gemini 3 ProGoogle | 1,485 | +/-4 | 42,072 | Proprietary | |
| 21 |  GPT-5.6 Sol (xhigh)OpenAI | 1,485 | +/-4 | 40,442 | OpenAI | Proprietary |
| 22 | GPT-6.1 Sol (Max)OpenAI | 1,484 | +/-8 | 6,320 | OpenAI | Proprietary |
| 23 | Gemini 3.6 Flash (High)Google | 1,483 | +/-4 | 36,122 | Proprietary | |
| 24 |  Qwen3.8-Max阿里巴巴 | 1,482 | +/-5 | 27,766 | 阿里巴巴 | Proprietary |
| 25 | Claude Opus 4.8 (High)Anthropic | 1,481 | +/-4 | 64,100 | Anthropic | Proprietary |
| 26 | GPT-5.5 (high)OpenAI | 1,481 | +/-4 | 69,401 | OpenAI | Proprietary |
| 27 | MiMo-V2.6-ProXiaomi | 1,479 | +/-9 | 4,075 | Xiaomi | MIT |
| 28 |  GLM-5.3 (max)智谱AI | 1,479 | +/-5 | 22,543 | 智谱AI | MIT |
| 29 | Gemini 3.5 Flash (High)Google | 1,477 | +/-4 | 48,486 | Proprietary | |
| 30 | GPT-6 Astra (max)OpenAI | 1,477 | +/-6 | 12,163 | OpenAI | Proprietary |
| 31 | GPT-5.5OpenAI | 1,477 | +/-4 | 70,985 | OpenAI | Proprietary |
| 32 | GLM-5.2 (max)智谱AI | 1,476 | +/-4 | 49,820 | 智谱AI | MIT |
| 33 | Gemini 3.5 Flash (medium)Google DeepMind | 1,476 | +/-4 | 47,292 | Google DeepMind | Proprietary |
| 34 | GPT-5.2 Chat (20260210)OpenAI | 1,476 | +/-4 | 35,999 | OpenAI | Proprietary |
| 35 | GPT-5.4 (High)OpenAI | 1,475 | +/-4 | 64,776 | OpenAI | Proprietary |
| 36 | Claude Sonnet 5.5 (xHigh)Anthropic | 1,475 | +/-8 | 6,831 | Anthropic | Proprietary |
| 37 |  Grok 4.20 Beta 1SpaceXAI | 1,475 | +/-5 | 27,908 | SpaceXAI | Proprietary |
| 38 |  DeepSeek-V4.1-Flash (Max)DeepSeek | 1,475 | +/-6 | 12,793 | DeepSeek | MIT |
| 39 | Qwen3.7-Max-Preview阿里巴巴 | 1,474 | +/-10 | 3,934 | 阿里巴巴 | Proprietary |
| 40 | Claude Opus 4.8Anthropic | 1,474 | +/-4 | 65,172 | Anthropic | Proprietary |
| 41 | GLM-5.3-FlashZ.ai | 1,474 | +/-5 | 32,925 | Z.ai | MIT |
| 42 | Claude Opus 4.5 (High, 32K, 20251101)Anthropic | 1,474 | +/-4 | 37,509 | Anthropic | Proprietary |
| 43 | GPT-5.5 InstantOpenAI | 1,473 | +/-5 | 27,158 | OpenAI | Proprietary |
| 44 | Gemini 3.0 FlashGoogle DeepMind | 1,473 | +/-4 | 31,325 | Google DeepMind | Proprietary |
| 45 | Claude Sonnet 4.6Anthropic | 1,472 | +/-4 | 70,931 | Anthropic | Proprietary |
| 46 | Grok 4.20 Beta (Reasoning, 0309)SpaceXAI | 1,472 | +/-4 | 66,436 | SpaceXAI | Proprietary |
| 47 | Grok 4.20 Multi-Agent Beta (0309)SpaceXAI | 1,470 | +/-4 | 64,946 | SpaceXAI | Proprietary |
| 48 | Claude Opus 4.5 (20251101)Anthropic | 1,470 | +/-3 | 72,899 | Anthropic | Proprietary |
| 49 | MiMo-V2.5-ProXiaomi | 1,467 | +/-4 | 75,721 | Xiaomi | MIT |
| 50 |  ERNIE 5.1Baidu | 1,467 | +/-5 | 39,468 | Baidu | Proprietary |
| 51 | GPT-5.6 Terra (xhigh)OpenAI | 1,466 | +/-4 | 43,271 | OpenAI | Proprietary |
| 52 | Grok 4.5xAI | 1,466 | +/-4 | 39,896 | xAI | Proprietary |
| 53 | Grok 4.1 (Thinking)SpaceXAI | 1,465 | +/-3 | 66,509 | SpaceXAI | Proprietary |
| 54 | GLM 5.1智谱AI | 1,465 | +/-4 | 62,889 | 智谱AI | MIT |
| 55 | GPT-5.4OpenAI | 1,465 | +/-4 | 67,870 | OpenAI | Proprietary |
| 56 | DeepSeek-V4-Pro (High, 20260813)DeepSeek | 1,465 | +/-7 | 9,800 | DeepSeek | MIT |
| 57 | Qwen3.5 Max Preview阿里巴巴 | 1,464 | +/-5 | 22,600 | 阿里巴巴 | Proprietary |
| 58 | Claude Sonnet 5 (High)Anthropic | 1,461 | +/-4 | 49,595 | Anthropic | Proprietary |
| 59 | Kimi K2.6Moonshot AI | 1,461 | +/-4 | 40,013 | Moonshot AI | Modified MIT |
| 60 | Qwen3.6-Max-Preview阿里巴巴 | 1,461 | +/-8 | 5,426 | 阿里巴巴 | Proprietary |
| 61 | GPT-6 Sol (Max)OpenAI | 1,459 | +/-6 | 12,466 | OpenAI | Proprietary |
| 62 | Grok 4.1SpaceXAI | 1,459 | +/-3 | 68,648 | SpaceXAI | Proprietary |
| 63 | Gemini 3 Flash (Minimal)Google | 1,458 | +/-3 | 91,039 | Proprietary | |
| 64 | DeepSeek-V4-ProDeepSeek-AI | 1,458 | +/-4 | 57,594 | DeepSeek-AI | MIT |
| 65 | GLM-5智谱AI | 1,457 | +/-4 | 29,272 | 智谱AI | MIT |
| 66 | Claude Sonnet 4.5 (High, 32K, 20250929)Anthropic | 1,456 | +/-3 | 83,940 | Anthropic | Proprietary |
| 67 |  Dola Seed 2.0 ProBytedance | 1,456 | +/-3 | 79,140 | Bytedance | Proprietary |
| 68 | GPT-5.1 (high)OpenAI | 1,456 | +/-4 | 41,651 | OpenAI | Proprietary |
| 69 | Qwen3.7-Plus阿里巴巴 | 1,456 | +/-4 | 42,075 | 阿里巴巴 | Proprietary |
| 70 | Claude Sonnet 4.5 (20250929)Anthropic | 1,455 | +/-3 | 82,688 | Anthropic | Proprietary |
| 71 | DeepSeek-V4-Pro Preview (High)DeepSeek | 1,455 | +/-4 | 54,902 | DeepSeek | MIT |
| 72 | Gemini 3.5 Flash-LiteGoogle DeepMind | 1,455 | +/-4 | 35,851 | Google DeepMind | Proprietary |
| 73 | Grok 4.6 (high)xAI | 1,454 | +/-5 | 28,965 | xAI | Proprietary |
| 74 |  Hy3腾讯AI实验室 | 1,453 | +/-6 | 12,208 | 腾讯AI实验室 | Apache 2.0 |
| 75 | GPT-5.6 Luna (xhigh)OpenAI | 1,453 | +/-4 | 41,109 | OpenAI | Proprietary |
| 76 | Gemma 4 31BGoogle | 1,452 | +/-7 | 6,147 | Apache 2.0 | |
| 77 | MiMo-V2.6-FlashXiaomi | 1,451 | +/-6 | 10,917 | Xiaomi | MIT |
| 78 | GPT-5.3 ChatOpenAI | 1,450 | +/-4 | 34,345 | OpenAI | Proprietary |
| 79 | Kimi K2.5 (thinking)Moonshot AI | 1,450 | +/-3 | 74,759 | Moonshot AI | Modified MIT |
| 80 | Claude Opus 4.1 (Thinking 16K, 20250805)Anthropic | 1,450 | +/-3 | 49,542 | Anthropic | Proprietary |
| 81 | ERNIE-5.0-Preview (1203)Baidu | 1,449 | +/-7 | 9,806 | Baidu | Proprietary |
| 82 | Claude Opus 4.1 (20250805)Anthropic | 1,448 | +/-3 | 77,238 | Anthropic | Proprietary |
| 83 | MiMo-V2-ProXiaomi | 1,448 | +/-5 | 25,447 | Xiaomi | Proprietary |
| 84 |  Step 5 PreviewStepFun | 1,447 | +/-8 | 7,116 | StepFun | Proprietary |
| 85 | GPT-5.4 mini (High)OpenAI | 1,447 | +/-4 | 63,449 | OpenAI | Proprietary |
| 86 | ERNIE 5.0 (0110)Baidu | 1,446 | +/-4 | 36,487 | Baidu | Proprietary |
| 87 | Gemini 2.5-ProGoogle DeepMind | 1,445 | +/-2 | 125,366 | Google DeepMind | Proprietary |
| 88 | GPT-4.5 Preview (2025-02-27)OpenAI | 1,445 | +/-6 | 14,547 | OpenAI | Proprietary |
| 89 | Grok 4.7 (xHigh)SpaceXAI | 1,444 | +/-6 | 11,591 | SpaceXAI | Proprietary |
| 90 | Qwen 3.6 Plus Preview阿里巴巴 | 1,443 | +/-4 | 48,165 | 阿里巴巴 | Proprietary |
| 91 | GPT-6 Luna (Max)OpenAI | 1,443 | +/-5 | 49,617 | OpenAI | Proprietary |
| 92 | ChatGPT-4o Latest (20250326)OpenAI | 1,443 | +/-3 | 81,914 | OpenAI | Proprietary |
| 93 | Grok 4.3 BetaxAI | 1,442 | +/-4 | 71,347 | xAI | Proprietary |
| 94 | Qwen3.5-397B-A17B阿里巴巴 | 1,442 | +/-3 | 93,567 | 阿里巴巴 | Apache 2.0 |
| 95 | GLM-4.7智谱AI | 1,441 | +/-6 | 12,253 | 智谱AI | MIT |
| 96 |  MiniMax M3MiniMaxAI | 1,441 | +/-4 | 63,973 | MiniMaxAI | MiniMax Community License |
| 97 | InklingThinking Machines Lab | 1,441 | +/-4 | 40,648 | Thinking Machines Lab | Apache 2.0 |
| 98 | DeepSeek-V4-Flash Preview (High)DeepSeek | 1,439 | +/-4 | 51,809 | DeepSeek | MIT |
| 99 | GPT-5.1OpenAI | 1,439 | +/-4 | 44,434 | OpenAI | Proprietary |
| 100 | Qwen3.8-27BAlibaba | 1,438 | +/-5 | 23,294 | Alibaba | Apache 2.0 |
| 101 | Gemma 4 26B A4BGoogle | 1,438 | +/-7 | 6,094 | Apache 2.0 | |
| 102 | GPT-5.2 (high)OpenAI | 1,437 | +/-4 | 49,655 | OpenAI | Proprietary |
| 103 | LongCat-Flash-Chat Exp (2602)Meituan | 1,436 | +/-5 | 29,291 | Meituan | Proprietary |
| 104 | DeepSeek-V4-FlashDeepSeek-AI | 1,436 | +/-4 | 51,987 | DeepSeek-AI | MIT |
| 105 | GPT-5.2OpenAI | 1,435 | +/-3 | 83,855 | OpenAI | Proprietary |
| 106 | GPT-5 (high)OpenAI | 1,435 | +/-5 | 31,722 | OpenAI | Proprietary |
| 107 | Qwen3 Max (Preview)阿里巴巴 | 1,434 | +/-5 | 27,370 | 阿里巴巴 | Proprietary |
| 108 | MiMo-V2.5Xiaomi | 1,434 | +/-4 | 47,362 | Xiaomi | MIT |
| 109 | GLM-5V-Turbo智谱AI | 1,434 | +/-6 | 10,482 | 智谱AI | Proprietary |
| 110 | Gemini 3.1 Flash-Lite PreviewGoogle | 1,433 | +/-4 | 64,211 | Proprietary | |
| 111 |  Mistral Large 4Mistral | 1,432 | +/-7 | 7,451 | Mistral | Proprietary |
| 112 | o3 (2025-04-16)OpenAI | 1,432 | +/-4 | 58,908 | OpenAI | Proprietary |
| 113 | MiMo-V2-OmniXiaomi | 1,431 | +/-6 | 20,619 | Xiaomi | Proprietary |
| 114 | Kimi K2.5 (Instant)Moonshot | 1,431 | +/-7 | 8,376 | Moonshot | Modified MIT |
| 115 | Grok 4.1 Fast (reasoning)xAI | 1,430 | +/-3 | 57,541 | xAI | Proprietary |
| 116 | Kimi K2 Thinking TurboMoonshot | 1,430 | +/-3 | 63,368 | Moonshot | Modified MIT |
| 117 | Mistral Medium 3.5MistralAI | 1,426 | +/-6 | 11,778 | MistralAI | Modified MIT |
| 118 | GPT-5 ChatOpenAI | 1,426 | +/-4 | 31,188 | OpenAI | Proprietary |
| 119 | Claude Opus 4 (Thinking 16K, 20250514)Anthropic | 1,426 | +/-4 | 35,904 | Anthropic | Proprietary |
| 120 |  amazon-nova-experimental-chat (26-02-10)Amazon | 1,426 | +/-10 | 3,526 | Amazon | Proprietary |
| 121 |  NVIDIA-Nemotron-3-Ultra-550B-A55B (NVFP4)Nvidia | 1,426 | +/-7 | 11,567 | Nvidia | OpenMDW-1.1 |
| 122 | DeepSeek-V3.2-Exp (Thinking)DeepSeek | 1,425 | +/-7 | 8,998 | DeepSeek | MIT |
| 123 | DeepSeek V3.2DeepSeek-AI | 1,425 | +/-3 | 48,112 | DeepSeek-AI | MIT |
| 124 | Muse GlimmerMeta | 1,425 | +/-10 | 3,897 | Meta | Apache-2.0 |
| 125 | GLM-4.6智谱AI | 1,424 | +/-4 | 35,824 | 智谱AI | MIT |
| 126 | Qwen3-Max (2025-09-23)Alibaba | 1,423 | +/-6 | 9,027 | Alibaba | Proprietary |
| 127 | DeepSeek-V3.2 (Thinking)DeepSeek | 1,422 | +/-4 | 42,071 | DeepSeek | MIT |
| 128 | Qwen3-235B-A22B-Instruct (2507)Alibaba | 1,422 | +/-3 | 98,134 | Alibaba | Apache 2.0 |
| 129 | DeepSeek V3.2-ExpDeepSeek-AI | 1,422 | +/-6 | 12,004 | DeepSeek-AI | MIT |
| 130 | DeepSeek-R1 (0528)DeepSeek | 1,421 | +/-6 | 18,167 | DeepSeek | MIT |
| 131 | Grok 4 Fast (Non-Reasoning)SpaceXAI | 1,419 | +/-8 | 6,340 | SpaceXAI | Proprietary |
| 132 | Kimi K2 Preview (0905)Moonshot | 1,419 | +/-7 | 11,589 | Moonshot | Modified MIT |
| 133 | ERNIE-5.0-Preview (1022)Baidu | 1,418 | +/-9 | 4,753 | Baidu | Proprietary |
| 134 | Kimi K2 Preview (0711)Moonshot | 1,418 | +/-5 | 27,146 | Moonshot | Modified MIT |
| 135 | DeepSeek-V3.1DeepSeek-AI | 1,417 | +/-6 | 14,610 | DeepSeek-AI | MIT |
| 136 | Qwen3.5-122B-A10B阿里巴巴 | 1,416 | +/-4 | 29,732 | 阿里巴巴 | Apache 2.0 |
| 137 | DeepSeek-V3.1-Terminus (Thinking)DeepSeek | 1,416 | +/-10 | 3,375 | DeepSeek | MIT |
| 138 | nvidia-nemotron-3.5-super-vl-120b-a12b-nvfp4Nvidia | 1,416 | +/-12 | 2,523 | Nvidia | OpenMDW-1.1 |
| 139 | DeepSeek-V3.1 (thinking)DeepSeek-AI | 1,416 | +/-7 | 11,459 | DeepSeek-AI | MIT |
| 140 | MiniMax-M2.7MiniMaxAI | 1,415 | +/-3 | 84,977 | MiniMaxAI | Modified MIT |
| 141 | DeepSeek-V3.1 TerminusDeepSeek-AI | 1,415 | +/-10 | 3,608 | DeepSeek-AI | MIT |
| 142 | GPT-4.1 (2025-04-14)OpenAI | 1,415 | +/-4 | 50,312 | OpenAI | Proprietary |
| 143 | Claude Opus 4 (20250514)Anthropic | 1,414 | +/-4 | 43,043 | Anthropic | Proprietary |
| 144 | Claude Haiku 4.5 (20251001)Anthropic | 1,414 | +/-2 | 149,199 | Anthropic | Proprietary |
| 145 | Mistral Large 3MistralAI | 1,414 | +/-3 | 84,205 | MistralAI | Apache 2.0 |
| 146 | Qwen3-VL-235B-A22B-Instruct阿里巴巴 | 1,413 | +/-7 | 11,320 | 阿里巴巴 | Apache 2.0 |
| 147 | amazon-nova-experimental-chat (26-01-10)Amazon | 1,413 | +/-10 | 3,539 | Amazon | Proprietary |
| 148 | Grok 3 Preview (02-24)SpaceXAI | 1,411 | +/-4 | 32,499 | SpaceXAI | Proprietary |
| 149 | Grok 4 (0709)SpaceXAI | 1,411 | +/-4 | 39,693 | SpaceXAI | Proprietary |
| 150 | GLM-4.5智谱AI | 1,411 | +/-5 | 23,796 | 智谱AI | MIT |
| 151 | Gemini 2.5 FlashGoogle DeepMind | 1,409 | +/-2 | 125,521 | Google DeepMind | Proprietary |
| 152 | Hy3 previewTencent | 1,409 | +/-7 | 6,879 | Tencent | tencent-hunyuan-community |
| 153 | Qwen3.5-27B阿里巴巴 | 1,409 | +/-4 | 28,671 | 阿里巴巴 | Apache 2.0 |
| 154 | Mistral Medium 3.1 (2508)Mistral | 1,408 | +/-3 | 95,321 | Mistral | Proprietary |
| 155 | Grok 4 Fast (reasoning)xAI | 1,405 | +/-5 | 18,269 | xAI | Proprietary |
| 156 | Inkling-SmallThinky | 1,404 | +/-5 | 32,304 | Thinky | Apache 2.0 |
| 157 | Gemini 2.5 Flash Preview (09-2025)Google | 1,403 | +/-4 | 32,995 | Proprietary | |
| 158 | o1 (2024-12-17)OpenAI | 1,402 | +/-4 | 27,807 | OpenAI | Proprietary |
| 159 | Qwen3-235B-A22B (Non-Thinking)Alibaba | 1,402 | +/-5 | 37,484 | Alibaba | Apache 2.0 |
| 160 | Claude Sonnet 4 (Thinking 32K, 20250514)Anthropic | 1,402 | +/-4 | 34,132 | Anthropic | Proprietary |
| 161 | GPT-5.4 nano (High)OpenAI | 1,401 | +/-4 | 62,396 | OpenAI | Proprietary |
| 162 | LongCat-Flash-ChatMeituan | 1,401 | +/-6 | 11,168 | Meituan | MIT |
| 163 | Qwen3-235B-A22B-Thinking (2507)Alibaba | 1,400 | +/-7 | 8,823 | Alibaba | Apache 2.0 |
| 164 | Qwen3-Next阿里巴巴 | 1,399 | +/-5 | 22,760 | 阿里巴巴 | Apache 2.0 |
| 165 | DeepSeek-R1DeepSeek-AI | 1,398 | +/-5 | 18,524 | DeepSeek-AI | MIT |
| 166 | Step 3.5 FlashStepFunAI | 1,396 | +/-4 | 61,802 | StepFunAI | Proprietary |
| 167 | DeepSeek-V3 (0324)DeepSeek | 1,395 | +/-4 | 44,973 | DeepSeek | MIT |
| 168 | Qwen3-VL-235B-A22B-Instruct (thinking)阿里巴巴 | 1,395 | +/-7 | 7,859 | 阿里巴巴 | Apache 2.0 |
| 169 | Qwen3.5-35B-A3B阿里巴巴 | 1,394 | +/-4 | 30,457 | 阿里巴巴 | Apache 2.0 |
| 170 | Hunyuan Vision 1.5 ThinkingTencent | 1,394 | +/-12 | 2,170 | Tencent | Proprietary |
| 171 | Step 3.5 FlashStepFunAI | 1,393 | +/-4 | 60,332 | StepFunAI | Apache 2.0 |
| 172 | amazon-nova-experimental-chat (12-10)Amazon | 1,393 | +/-9 | 3,733 | Amazon | Proprietary |
| 173 | MiMo-V2-Flash (Non-Thinking)Xiaomi | 1,392 | +/-3 | 48,088 | Xiaomi | MIT |
| 174 | o4-mini (2025-04-16)OpenAI | 1,391 | +/-4 | 44,855 | OpenAI | Proprietary |
| 175 | Claude Sonnet 4 (20250514)Anthropic | 1,391 | +/-4 | 39,163 | Anthropic | Proprietary |
| 176 | MiniMax M2.5MiniMaxAI | 1,391 | +/-4 | 42,815 | MiniMaxAI | Modified MIT |
| 177 | GPT-5-mini (high)OpenAI | 1,390 | +/-5 | 26,804 | OpenAI | Proprietary |
| 178 | OpenAI o1OpenAI | 1,389 | +/-5 | 31,122 | OpenAI | Proprietary |
| 179 | Claude 3.7 Sonnet (Thinking 32K, 20250219)Anthropic | 1,388 | +/-4 | 38,159 | Anthropic | Proprietary |
| 180 | Qwen3-Coder-480B-A35B阿里巴巴 | 1,388 | +/-5 | 25,195 | 阿里巴巴 | Apache 2.0 |
| 181 |  Phi-3-medium 14B-preview微软 | 1,387 | +/-5 | 32,782 | 微软 | Proprietary |
| 182 | Hunyuan T1 (20250711)Tencent | 1,387 | +/-9 | 4,595 | Tencent | Proprietary |
| 183 | Solar Pro 4Upstage | 1,386 | +/-5 | 21,281 | Upstage | Proprietary |
| 184 | MiMo-V2-Flash (Thinking)Xiaomi | 1,386 | +/-6 | 11,142 | Xiaomi | MIT |
| 185 | MiniMax-M2.1 PreviewMiniMax | 1,384 | +/-5 | 17,239 | MiniMax | MIT |
| 186 | Hunyuan TurboS (20250416)Tencent | 1,383 | +/-6 | 10,626 | Tencent | Proprietary |
| 187 | GPT-4.1 mini (2025-04-14)OpenAI | 1,383 | +/-4 | 38,709 | OpenAI | Proprietary |
| 188 | Qwen3-30B-A3B-Instruct (2507)Alibaba | 1,382 | +/-5 | 23,322 | Alibaba | Apache 2.0 |
| 189 | Gemini 2.5 Flash-Lite Preview (Non-Thinking, 09-2025)Google | 1,379 | +/-3 | 47,437 | Proprietary | |
| 190 | Trinity Large Preview Apache 2.0 | 1,379 | +/-4 | 31,266 | — | — |
| 191 | GLM-4.6V智谱AI | 1,378 | +/-11 | 2,837 | 智谱AI | MIT |
| 192 | Qwen3-235B-A22B阿里巴巴 | 1,375 | +/-5 | 25,948 | 阿里巴巴 | Apache 2.0 |
| 193 | Gemini 2.5 Flash-Lite Preview (Thinking, 06-17)Google | 1,375 | +/-5 | 32,381 | Proprietary | |
| 194 | Claude 3.5 Sonnet (20241022)Anthropic | 1,374 | +/-3 | 87,769 | Anthropic | Proprietary |
| 195 | Qwen2.5-Max阿里巴巴 | 1,374 | +/-4 | 32,487 | 阿里巴巴 | Proprietary |
| 196 | GLM-4.5-Air智谱AI | 1,373 | +/-4 | 30,660 | 智谱AI | MIT |
| 197 | Claude 3.7 Sonnet (20250219)Anthropic | 1,373 | +/-4 | 42,572 | Anthropic | Proprietary |
| 198 | Qwen3-Next (thinking)阿里巴巴 | 1,369 | +/-6 | 13,474 | 阿里巴巴 | Apache 2.0 |
| 199 | Trinity Large Thinking Apache 2.0 | 1,367 | +/-5 | 30,337 | — | — |
| 200 | Gemma 3 27B ITGoogle | 1,365 | +/-4 | 46,666 | Gemma | |
| 201 | GLM-4.7-Flash智谱AI | 1,365 | +/-6 | 12,021 | 智谱AI | MIT |
| 202 | MiniMax-M1MiniMax | 1,364 | +/-4 | 34,303 | MiniMax | Apache 2.0 |
| 203 | amazon-nova-experimental-chat (11-10)Amazon | 1,364 | +/-4 | 26,039 | Amazon | Proprietary |
| 204 | Grok 3 Mini (High)SpaceXAI | 1,364 | +/-5 | 15,596 | SpaceXAI | Proprietary |
| 205 | o3-mini (High)OpenAI | 1,364 | +/-5 | 18,589 | OpenAI | Proprietary |
| 206 | NVIDIA-Nemotron-3-Super-120B-A12BNvidia | 1,360 | +/-7 | 7,878 | Nvidia | NVIDIA Open Model |
| 207 | Gemini 2.0 Flash (001)Google | 1,360 | +/-4 | 43,446 | Proprietary | |
| 208 | DeepSeek-V3DeepSeek-AI | 1,358 | +/-5 | 21,770 | DeepSeek-AI | DeepSeek |
| 209 | Grok 3 Mini BetaSpaceXAI | 1,358 | +/-5 | 20,931 | SpaceXAI | Proprietary |
| 210 | Mistral Small 3.2 (2506)Mistral | 1,356 | +/-5 | 17,462 | Mistral | Apache 2.0 |
| 211 | INTELLECT-3 MIT | 1,356 | +/-8 | 5,431 | — | — |
| 212 |  Command A (03-2025)Cohere | 1,354 | +/-3 | 55,766 | Cohere | CC-BY-NC-4.0 |
| 213 | Gemini 2.0 Flash-Lite Preview (02-05)Google | 1,354 | +/-4 | 24,955 | Proprietary | |
| 214 | GLM-4.5V智谱AI | 1,352 | +/-8 | 4,821 | 智谱AI | MIT |
| 215 | GPT OSS 120BOpenAI | 1,352 | +/-4 | 30,206 | OpenAI | Apache 2.0 |
| 216 | NVIDIA-Nemotron-3.5-Lightning-30B-A3B (NVFP4)Nvidia | 1,351 | +/-6 | 19,938 | Nvidia | OpenMDW-1.1 |
| 217 | Gemini 1.5 Pro (002)Google | 1,351 | +/-3 | 55,606 | Proprietary | |
| 218 | Step3StepFunAI | 1,349 | +/-8 | 6,293 | StepFunAI | Apache 2.0 |
| 219 | Hunyuan TurboS (20250226)Tencent | 1,349 | +/-12 | 2,220 | Tencent | Proprietary |
| 220 | OpenAI o3-miniOpenAI | 1,348 | +/-4 | 56,741 | OpenAI | Proprietary |
| 221 | amazon-nova-experimental-chat (10-20)Amazon | 1,348 | +/-6 | 11,626 | Amazon | Proprietary |
| 222 | Llama 3.1 Nemotron Ultra 253B v1Nvidia | 1,348 | +/-12 | 2,549 | Nvidia | Nvidia Open Model |
| 223 | Qwen3-32B阿里巴巴 | 1,347 | +/-9 | 3,926 | 阿里巴巴 | Apache 2.0 |
| 224 | amazon-nova-experimental-chat (10-09)Amazon | 1,346 | +/-11 | 2,781 | Amazon | Proprietary |
| 225 | GPT-4o (2024-05-13)OpenAI | 1,346 | +/-3 | 112,881 | OpenAI | Proprietary |
| 226 | Qwen-Plus (0125)Alibaba | 1,346 | +/-8 | 5,819 | Alibaba | Proprietary |
| 227 | MiniMax M2MiniMaxAI | 1,343 | +/-8 | 6,856 | MiniMaxAI | Apache 2.0 |
| 228 | Claude 3.5 Sonnet (20240620)Anthropic | 1,343 | +/-3 | 82,419 | Anthropic | Proprietary |
| 229 | Ling-flash-2.0 AntGroup | 1,343 | +/-7 | 6,947 | Group | MIT |
| 230 | Mercury 2 InceptionAI | 1,343 | +/-11 | 3,197 | AI | Proprietary |
| 231 | Llama 3.3 Nemotron Super 49B v1.5Nvidia | 1,343 | +/-10 | 3,328 | Nvidia | Nvidia Open |
| 232 | GLM-4-Plus (0111)Z.ai | 1,343 | +/-8 | 5,760 | Z.ai | Proprietary |
| 233 | Gemma 3 12B ITGoogle | 1,342 | +/-10 | 3,829 | Gemma | |
| 234 | Hunyuan Turbo (0110)Tencent | 1,341 | +/-12 | 2,290 | Tencent | Proprietary |
| 235 | Granite 4.2 30BIBM | 1,340 | +/-10 | 3,401 | IBM | Apache 2.0 |
| 236 | GPT-5 nano (High)OpenAI | 1,338 | +/-7 | 8,132 | OpenAI | Proprietary |
| 237 | OpenAI o1-miniOpenAI | 1,337 | +/-4 | 51,981 | OpenAI | Proprietary |
| 238 | Gemini Advanced (0514)Google | 1,336 | +/-5 | 50,148 | Proprietary | |
| 239 | Grok 2 (2024-08-13)SpaceXAI | 1,336 | +/-4 | 63,498 | SpaceXAI | Proprietary |
| 240 | GPT-4o (2024-08-06)OpenAI | 1,336 | +/-4 | 45,499 | OpenAI | Proprietary |
| 241 | QwQ-32B阿里巴巴 | 1,336 | +/-4 | 25,178 | 阿里巴巴 | Apache 2.0 |
| 242 | Llama 3.1 405B Instruct (BF16)Meta | 1,335 | +/-4 | 41,375 | Meta | Llama 3.1 Community |
| 243 | Amazon Nova 2 LiteAmazon | 1,335 | +/-6 | 12,413 | Amazon | Proprietary |
| 244 | Llama 3.1 405B Instruct (FP8)Meta | 1,334 | +/-4 | 59,656 | Meta | Llama 3.1 Community |
| 245 | Step-2 16K Exp (202412)StepFun | 1,333 | +/-9 | 4,833 | StepFun | Proprietary |
| 246 | Olmo 3.1 32B InstructAi2 | 1,329 | +/-6 | 11,817 | Ai2 | Apache 2.0 |
| 247 | Yi Lightning Proprietary | 1,328 | +/-5 | 27,332 | — | — |
| 248 | Llama 3.3 Nemotron Super 49B v1Nvidia | 1,328 | +/-12 | 2,218 | Nvidia | Nvidia |
| 249 | Llama 4 Maverick 17B 128E InstructMeta | 1,327 | +/-4 | 39,450 | Meta | Llama 4 |
| 250 | Qwen3-30B-A3B阿里巴巴 | 1,327 | +/-5 | 26,157 | 阿里巴巴 | Apache 2.0 |
| 251 | Hunyuan Large (2025-02-10)Tencent | 1,326 | +/-10 | 3,738 | Tencent | Proprietary |
| 252 | Claude 3.5 Haiku (20241022)Anthropic | 1,325 | +/-3 | 69,061 | Anthropic | Proprietary |
| 253 | GPT-4 Turbo (2024-04-09)OpenAI | 1,325 | +/-4 | 98,114 | OpenAI | Proprietary |
| 254 | Gemini 1.5 Pro (001)Google | 1,324 | +/-4 | 79,138 | Proprietary | |
| 255 | DeepSeek-V2.5 (1210)DeepSeek | 1,323 | +/-8 | 6,795 | DeepSeek | DeepSeek |
| 256 | Ring-flash-2.0 AntGroup | 1,323 | +/-7 | 6,990 | Group | MIT |
| 257 | Claude 3 Opus (20240229)Anthropic | 1,322 | +/-3 | 194,909 | Anthropic | Proprietary |
| 258 | GPT-4.1 nano (2025-04-14)OpenAI | 1,322 | +/-8 | 6,103 | OpenAI | Proprietary |
| 259 | Molmo 2 8BAi2 | 1,322 | +/-21 | 822 | Ai2 | Apache 2.0 |
| 260 | Llama 4 Scout 17B 16E InstructMeta | 1,322 | +/-5 | 29,826 | Meta | Llama |
| 261 | Step-1o Turbo (202506)StepFun | 1,320 | +/-7 | 8,901 | StepFun | Proprietary |
| 262 | GLM-4-PlusZhipu AI | 1,319 | +/-5 | 26,126 | Zhipu AI | Proprietary |
| 263 | Qwen-Max (0919)Alibaba | 1,318 | +/-6 | 16,478 | Alibaba | Qwen |
| 264 | Llama 3.3 70B InstructMeta | 1,318 | +/-4 | 54,497 | Meta | Llama-3.3 |
| 265 | GPT-4o mini (2024-07-18)OpenAI | 1,318 | +/-4 | 68,705 | OpenAI | Proprietary |
| 266 | Gemma 3n E4B ITGoogle | 1,318 | +/-5 | 22,137 | Gemma | |
| 267 | GPT OSS 20BOpenAI | 1,317 | +/-6 | 10,419 | OpenAI | Apache 2.0 |
| 268 | Qwen2.5-Plus (1127)Alibaba | 1,315 | +/-6 | 10,187 | Alibaba | Proprietary |
| 269 | Mistral Large 2 (2407)Mistral | 1,314 | +/-4 | 45,459 | Mistral | Mistral Research |
| 270 | Athene V2 Chat NexusFlow | 1,314 | +/-5 | 24,739 | — | — |
| 271 | NVIDIA-Nemotron-3-Nano-30B-A3B (BF16)Nvidia | 1,314 | +/-5 | 15,770 | Nvidia | NVIDIA Open Model |
| 272 | GPT-4 Turbo Preview (0125)OpenAI | 1,313 | +/-4 | 93,439 | OpenAI | Proprietary |
| 273 | GPT-4 Turbo Preview (1106)OpenAI | 1,313 | +/-4 | 100,105 | OpenAI | Proprietary |
| 274 | Hunyuan Standard (2025-02-10)Tencent | 1,311 | +/-10 | 3,904 | Tencent | Proprietary |
| 275 | Gemini 1.5 Flash (002)Google | 1,309 | +/-4 | 34,902 | Proprietary | |
| 276 | Grok 2 mini (2024-08-13)SpaceXAI | 1,308 | +/-4 | 52,567 | SpaceXAI | Proprietary |
| 277 | DeepSeek V2.5DeepSeek-AI | 1,307 | +/-5 | 24,572 | DeepSeek-AI | DeepSeek |
| 278 | Athene 70B (0725) CC-BY-NC-4.0 | 1,307 | +/-6 | 19,621 | — | — |
| 279 | Olmo 3 32B ThinkAi2 | 1,306 | +/-8 | 5,831 | Ai2 | Apache 2.0 |
| 280 | Mistral Large (2411)Mistral | 1,306 | +/-4 | 28,073 | Mistral | MRL |
| 281 | Mercury InceptionAI | 1,305 | +/-14 | 1,965 | AI | Proprietary |
| 282 | Magistral Medium (2506)Mistral | 1,305 | +/-6 | 11,510 | Mistral | Proprietary |
| 283 | Granite 4.1 8BIBM | 1,304 | +/-10 | 4,180 | IBM | Apache 2.0 |
| 284 | Gemma 3 4B ITGoogle | 1,303 | +/-9 | 4,171 | Gemma | |
| 285 | Qwen2.5-VL-72B-Instruct阿里巴巴 | 1,303 | +/-4 | 39,406 | 阿里巴巴 | Qwen |
| 286 | Mistral Small 3.1 24B Instruct (2503)Mistral | 1,303 | +/-5 | 32,727 | Mistral | Apache 2.0 |
| 287 | Llama 3.1 Nemotron 70B InstructNvidia | 1,299 | +/-8 | 7,140 | Nvidia | Llama 3.1 |
| 288 | Hunyuan Large VisionTencent | 1,294 | +/-9 | 5,279 | Tencent | Proprietary |
| 289 | Llama 3.1 70B InstructMeta | 1,293 | +/-4 | 55,240 | Meta | Llama 3.1 Community |
| 290 | Granite 4.2 8BIBM | 1,290 | +/-11 | 3,254 | IBM | Apache 2.0 |
| 291 | Amazon Nova Pro 1.0Amazon | 1,290 | +/-5 | 24,745 | Amazon | Proprietary |
| 292 | Granite 4.2 3BIBM | 1,290 | +/-12 | 3,251 | IBM | Apache 2.0 |
| 293 | Gemma 2 27B ITGoogle | 1,290 | +/-3 | 75,754 | Gemma license | |
| 294 | Jamba 1.5 Large Jamba Open | 1,289 | +/-7 | 8,662 | — | — |
| 295 | Reka Core (20240904) Proprietary | 1,288 | +/-7 | 7,312 | — | — |
| 296 | GPT-4 (0314)OpenAI | 1,288 | +/-5 | 54,173 | OpenAI | Proprietary |
| 297 | Granite-4.0-H-SmallIBM | 1,287 | +/-8 | 5,630 | IBM | Apache 2.0 |
| 298 | Llama 3.1 Nemotron 51B InstructNvidia | 1,287 | +/-10 | 3,749 | Nvidia | Llama 3.1 |
| 299 | Olmo 3.1 32B ThinkAi2 | 1,287 | +/-7 | 8,301 | Ai2 | Apache 2.0 |
| 300 | Gemini 1.5 Flash (001)Google | 1,287 | +/-5 | 62,833 | Proprietary | |
| 301 | Llama 3.1 Tulu 3 70BAi2 | 1,286 | +/-10 | 2,846 | Ai2 | Llama 3.1 |
| 302 | Claude 3 Sonnet (20240229)Anthropic | 1,281 | +/-4 | 109,284 | Anthropic | Proprietary |
| 303 | Gemma 2 9B IT SimPO MIT | 1,280 | +/-7 | 10,072 | — | — |
| 304 | Nemotron-4 340B InstructNvidia | 1,277 | +/-5 | 19,659 | Nvidia | NVIDIA Open Model |
| 305 | Llama 3 70B InstructMeta | 1,277 | +/-4 | 156,876 | Meta | Llama 3 Community |
| 306 | Command R+ (08-2024)Cohere | 1,276 | +/-7 | 9,866 | Cohere | CC-BY-NC-4.0 |
| 307 | GPT-4 (0613)OpenAI | 1,276 | +/-4 | 88,723 | OpenAI | Proprietary |
| 308 | Mistral Small 24B Instruct (2501)Mistral | 1,274 | +/-6 | 14,681 | Mistral | Apache 2.0 |
| 309 | GLM-4 (0520)Z.ai | 1,273 | +/-7 | 9,788 | Z.ai | Proprietary |
| 310 | Reka Flash (20240904) Proprietary | 1,272 | +/-7 | 7,536 | — | — |
| 311 | Qwen2.5-Coder-32B-Instruct阿里巴巴 | 1,271 | +/-8 | 5,432 | 阿里巴巴 | Apache 2.0 |
| 312 | Aya Expanse 32BCohere | 1,267 | +/-5 | 27,124 | Cohere | CC-BY-NC-4.0 |
| 313 | Gemma 2 9B ITGoogle | 1,267 | +/-4 | 54,611 | Gemma license | |
| 314 | DeepSeek-Coder-V2DeepSeek | 1,265 | +/-6 | 15,147 | DeepSeek | DeepSeek License |
| 315 | Qwen2-72B-Instruct阿里巴巴 | 1,262 | +/-5 | 37,325 | 阿里巴巴 | Qianwen LICENSE |
| 316 | Command R+Cohere | 1,262 | +/-4 | 77,554 | Cohere | CC-BY-NC-4.0 |
| 317 | Claude 3 Haiku (20240307)Anthropic | 1,262 | +/-4 | 117,701 | Anthropic | Proprietary |
| 318 | Amazon Nova Lite 1.0Amazon | 1,260 | +/-5 | 19,372 | Amazon | Proprietary |
| 319 | Gemini 1.5 Flash-8B (001)Google | 1,259 | +/-4 | 35,558 | Proprietary | |
| 320 | Phi-4-reasoning微软 | 1,256 | +/-5 | 24,126 | 微软 | MIT |
| 321 | OLMo 2 32B Instruct (0325)Ai2 | 1,252 | +/-11 | 3,334 | Ai2 | Apache-2.0 |
| 322 | Command R (08-2024)Cohere | 1,251 | +/-7 | 10,140 | Cohere | CC-BY-NC-4.0 |
| 323 | Mistral Large (2402)Mistral | 1,243 | +/-5 | 62,436 | Mistral | Proprietary |
| 324 | Amazon Nova Micro 1.0Amazon | 1,241 | +/-5 | 19,364 | Amazon | Proprietary |
| 325 | Jamba 1.5 Mini Jamba Open | 1,240 | +/-7 | 8,858 | — | — |
| 326 | Ministral 8B (2410)Mistral | 1,238 | +/-9 | 4,781 | Mistral | MRL |
| 327 | Gemini Pro (Dev API)Google | 1,237 | +/-7 | 18,354 | Proprietary | |
| 328 | Qwen1.5-110B-Chat阿里巴巴 | 1,234 | +/-6 | 26,195 | 阿里巴巴 | Qianwen LICENSE |
| 329 | Reka Flash 21B (Online, 20240226) Proprietary | 1,234 | +/-7 | 15,450 | — | — |
| 330 | Qwen1.5-72B-Chat阿里巴巴 | 1,234 | +/-5 | 39,302 | 阿里巴巴 | Qianwen LICENSE |
| 331 | Hunyuan Standard 256KTencent | 1,233 | +/-12 | 2,728 | Tencent | Proprietary |
| 332 | Mixtral 8x22B Instruct v0.1Mistral | 1,230 | +/-5 | 51,416 | Mistral | Apache 2.0 |
| 333 | Command RCohere | 1,227 | +/-5 | 54,036 | Cohere | CC-BY-NC-4.0 |
| 334 | Reka Flash 21B (20240226) Proprietary | 1,227 | +/-6 | 24,806 | — | — |
| 335 | GPT-3.5 Turbo (0125)OpenAI | 1,226 | +/-5 | 66,207 | OpenAI | Proprietary |
| 336 | Gemini ProGoogle | 1,224 | +/-12 | 6,390 | Proprietary | |
| 337 | Llama 3 8B InstructMeta | 1,224 | +/-4 | 104,642 | Meta | Llama 3 Community |
| 338 | Aya Expanse 8BCohere | 1,223 | +/-7 | 9,818 | Cohere | CC-BY-NC-4.0 |
| 339 | Mistral MediumMistral | 1,223 | +/-5 | 34,550 | Mistral | Proprietary |
| 340 | Llama 3.1 Tulu 3 8BAi2 | 1,221 | +/-11 | 2,896 | Ai2 | Llama 3.1 |
| 341 | Zephyr ORPO 141B A35B v0.1 Apache 2.0 | 1,213 | +/-11 | 4,652 | — | — |
| 342 | Yi 1.5 34B Chat Apache-2.0 | 1,213 | +/-5 | 24,146 | — | — |
| 343 | Llama 3.1 8B InstructMeta | 1,211 | +/-4 | 49,605 | Meta | Llama 3.1 Community |
| 344 | Granite 3.1 8B InstructIBM | 1,209 | +/-11 | 3,090 | IBM | Apache 2.0 |
| 345 | GPT-3.5 Turbo (1106)OpenAI | 1,204 | +/-9 | 16,619 | OpenAI | Proprietary |
| 346 | Qwen1.5-32B-ChatAlibaba | 1,204 | +/-6 | 21,741 | Alibaba | Qianwen LICENSE |
| 347 | Gemma 2 2B ITGoogle | 1,200 | +/-4 | 46,616 | Gemma license | |
| 348 |  Phi-3 Medium 4K InstructMicrosoft | 1,198 | +/-5 | 25,055 | Microsoft | MIT |
| 349 | Mixtral 8x7B Instruct v0.1Mistral | 1,197 | +/-4 | 73,503 | Mistral | Apache 2.0 |
| 350 | DBRX Instruct Preview DBRX LICENSE | 1,196 | +/-6 | 32,191 | — | — |
| 351 | Qwen1.5-14B-Chat阿里巴巴 | 1,191 | +/-7 | 17,839 | 阿里巴巴 | Qianwen LICENSE |
| 352 | InternLM2.5 20B Chat Other | 1,191 | +/-7 | 9,901 | — | — |
| 353 | DeepSeek LLM 67B ChatDeepSeek | 1,185 | +/-11 | 4,932 | DeepSeek | DeepSeek License |
| 354 | WizardLM 70BMicrosoft | 1,185 | +/-9 | 8,214 | Microsoft | Llama 2 Community |
| 355 | Yi 34B Chat Yi License | 1,184 | +/-7 | 15,483 | — | — |
| 356 | Granite 3.0 8B InstructIBM | 1,183 | +/-9 | 6,638 | IBM | Apache 2.0 |
| 357 | OpenChat 3.5 Apache-2.0 | 1,183 | +/-10 | 7,968 | — | — |
| 358 | OpenChat 3.5 (0106) Apache-2.0 | 1,183 | +/-8 | 12,637 | — | — |
| 359 | Gemma 1.1 7B ITGoogle | 1,183 | +/-6 | 23,893 | Gemma license | |
| 360 | Snowflake Arctic Instruct Apache 2.0 | 1,180 | +/-6 | 32,832 | — | — |
| 361 | Granite 3.1 2B InstructIBM | 1,179 | +/-11 | 3,188 | IBM | Apache 2.0 |
| 362 | Tulu 2 DPO 70B AI2 ImpACT Low-risk | 1,178 | +/-10 | 6,535 | — | — |
| 363 | OpenHermes 2.5 Mistral 7B Apache-2.0 | 1,176 | +/-10 | 5,006 | — | — |
| 364 | Vicuna 33B Non-commercial | 1,173 | +/-6 | 22,479 | — | — |
| 365 | Phi-3 Small 8K InstructMicrosoft | 1,171 | +/-6 | 17,766 | Microsoft | MIT |
| 366 | Starling LM 7B Beta Apache-2.0 | 1,171 | +/-7 | 16,056 | — | — |
| 367 | Llama 2 70B ChatMeta | 1,171 | +/-5 | 38,492 | Meta | Llama 2 Community |
| 368 | Starling LM 7B Alpha CC-BY-NC-4.0 | 1,168 | +/-8 | 10,224 | — | — |
| 369 | Llama 3.2 3B InstructMeta | 1,167 | +/-8 | 7,936 | Meta | Llama 3.2 |
| 370 | Nous Hermes 2 Mixtral 8x7B DPO Apache-2.0 | 1,165 | +/-12 | 3,777 | — | — |
| 371 | Granite 3.0 2B InstructIBM | 1,157 | +/-8 | 6,837 | IBM | Apache 2.0 |
| 372 | Llama 2 70B SteerLM ChatNvidia | 1,155 | +/-13 | 3,585 | Nvidia | Llama 2 Community |
| 373 | QwQ-32B-Preview阿里巴巴 | 1,154 | +/-11 | 3,231 | 阿里巴巴 | Apache 2.0 |
| 374 | SOLAR 10.7B Instruct v1.0Upstage | 1,153 | +/-13 | 4,155 | Upstage | CC-BY-NC-4.0 |
| 375 | Dolphin 2.2.1 Mistral 7B Apache-2.0 | 1,152 | +/-15 | 1,679 | — | — |
| 376 | MPT 30B Chat CC-BY-NC-SA-4.0 | 1,151 | +/-12 | 2,572 | — | — |
| 377 | WizardLM 13BMicrosoft | 1,150 | +/-9 | 7,044 | Microsoft | Llama 2 Community |
| 378 | Mistral 7B Instruct v0.2Mistral | 1,150 | +/-7 | 19,402 | Mistral | Apache-2.0 |
| 379 | Falcon 180B Chat Falcon-180B TII License | 1,148 | +/-17 | 1,295 | — | — |
| 380 | Qwen1.5-7B-Chat阿里巴巴 | 1,144 | +/-10 | 4,737 | 阿里巴巴 | Qianwen LICENSE |
| 381 | Phi-3 Mini 4K Instruct (June 2024)Microsoft | 1,143 | +/-6 | 12,297 | Microsoft | MIT |
| 382 | Vicuna 13B Llama 2 Community | 1,142 | +/-7 | 19,367 | — | — |
| 383 | Llama 2 13B ChatMeta | 1,142 | +/-7 | 19,174 | Meta | Llama 2 Community |
| 384 | Qwen-14B-Chat阿里巴巴 | 1,139 | +/-11 | 4,964 | 阿里巴巴 | Qianwen LICENSE |
| 385 | PaLM 2Google | 1,139 | +/-9 | 8,554 | Proprietary | |
| 386 | Gemma 7B ITGoogle | 1,138 | +/-9 | 8,925 | Gemma license | |
| 387 | Code Llama 34B InstructMeta | 1,137 | +/-9 | 7,366 | Meta | Llama 2 Community |
| 388 | Zephyr 7B Beta MIT | 1,131 | +/-9 | 11,118 | — | — |
| 389 | Phi-3 Mini 128K InstructMicrosoft | 1,130 | +/-7 | 20,685 | Microsoft | MIT |
| 390 | Phi-3 Mini 4K InstructMicrosoft | 1,128 | +/-6 | 20,118 | Microsoft | MIT |
| 391 | Guanaco 33B Non-commercial | 1,128 | +/-12 | 2,921 | — | — |
| 392 | Zephyr 7B Alpha MIT | 1,127 | +/-16 | 1,785 | — | — |
| 393 | StripedHyena Nous 7B Apache 2.0 | 1,122 | +/-11 | 5,182 | — | — |
| 394 | Code Llama 70B InstructMeta | 1,119 | +/-18 | 1,143 | Meta | Llama 2 Community |
| 395 | Gemma 1.1 2B ITGoogle | 1,117 | +/-8 | 10,854 | Gemma license | |
| 396 | Vicuna 7B Llama 2 Community | 1,115 | +/-9 | 6,923 | — | — |
| 397 | SmolLM2 1.7B Instruct Apache 2.0 | 1,115 | +/-14 | 2,199 | — | — |
| 398 | Llama 3.2 1B InstructMeta | 1,111 | +/-8 | 8,045 | Meta | Llama 3.2 |
| 399 | Mistral 7B InstructMistral | 1,110 | +/-9 | 8,977 | Mistral | Apache 2.0 |
| 400 | Llama 2 7B ChatMeta | 1,108 | +/-7 | 14,148 | Meta | Llama 2 Community |
| 401 | Gemma 2B ITGoogle | 1,094 | +/-11 | 4,780 | Gemma license | |
| 402 | Qwen1.5-4B-Chat阿里巴巴 | 1,091 | +/-9 | 7,597 | 阿里巴巴 | Qianwen LICENSE |
| 403 | OLMo 7B InstructAi2 | 1,074 | +/-11 | 6,328 | Ai2 | Apache-2.0 |
| 404 | Koala 13B Non-commercial | 1,071 | +/-10 | 6,965 | — | — |
| 405 | Alpaca 13B Non-commercial | 1,070 | +/-11 | 5,745 | — | — |
| 406 | GPT4All 13B Snoozy Non-commercial | 1,068 | +/-15 | 1,743 | — | — |
| 407 | MPT 7B Chat CC-BY-NC-SA-4.0 | 1,064 | +/-12 | 3,924 | — | — |
| 408 | ChatGLM3 6B Apache-2.0 | 1,057 | +/-12 | 4,658 | — | — |
| 409 | RWKV-4 Raven 14B Apache 2.0 | 1,042 | +/-11 | 4,845 | — | — |
| 410 | ChatGLM2 6B Apache-2.0 | 1,025 | +/-14 | 2,658 | — | — |
| 411 | OpenAssistant Pythia 12B Apache 2.0 | 1,024 | +/-11 | 6,310 | — | — |
| 412 | ChatGLM 6B Non-commercial | 996 | +/-12 | 4,914 | — | — |
| 413 | FastChat-T5 3B Apache 2.0 | 993 | +/-12 | 4,203 | — | — |
| 414 | Dolly 2.0 12B MIT | 983 | +/-13 | 3,412 | — | — |
| 415 | LLaMA 13BMeta | 975 | +/-16 | 2,391 | Meta | Non-commercial |
| 416 | StableLM Tuned Alpha 7B CC-BY-NC-SA-4.0 | 953 | +/-13 | 3,287 | — | — |
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。