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
Opus 4.7 (thinking)
最高得分
1,503
模型数量
200
数据版本
2026年04月24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。
筛选条件
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 |
|---|---|---|---|---|---|---|
| Opus 4.7 (thinking) | 1,503 | +/-8 | 5,321 | Anthropic | Proprietary | |
| Claude Opus 4.6 (thinking) | 1,503 | +/-5 | 20,192 | Anthropic | Proprietary | |
| Claude Opus 4.6 | 1,496 | +/-5 | 21,537 | Anthropic | Proprietary | |
| 4 | Opus 4.7 | 1,494 | +/-8 | 6,017 | Anthropic | Proprietary |
| 5 | Gemini 3.1 Pro Preview | 1,493 | +/-5 | 25,353 | Google Deep Mind | Proprietary |
| 6 | Muse Spark | 1,492 | +/-7 | 7,213 | Facebook AI研究实验室 | Proprietary |
| 7 | Gemini 3.0 Pro (Preview 11-2025) | 1,486 | +/-4 | 41,383 | Google Deep Mind | Proprietary |
| 8 | grok-4.20-beta1 | 1,482 | +/-6 | 14,620 | xAI | Proprietary |
| 9 | gpt-5.4-high | 1,481 | +/-6 | 13,593 | OpenAI | Proprietary |
| 10 | grok-4.20-beta-0309-reasoning | 1,479 | +/-6 | 13,841 | xAI | Proprietary |
| 11 | gpt-5.2-chat-latest-20260210 | 1,476 | +/-5 | 19,964 | OpenAI | Proprietary |
| 12 | grok-4.20-multi-agent-beta-0309 | 1,476 | +/-5 | 14,223 | xAI | Proprietary |
| 13 | Gemini 3.0 Flash | 1,474 | +/-4 | 30,791 | Google Deep Mind | Proprietary |
| 14 | Claude Opus 4 (thinking-32k) | 1,473 | +/-4 | 37,164 | Anthropic | Proprietary |
| 15 | GLM 5.1 | 1,470 | +/-7 | 9,028 | 智谱AI | MIT |
| 16 | Grok 4.1 Thinking | 1,469 | +/-4 | 51,489 | xAI | Proprietary |
| 17 | Claude Opus 4 | 1,469 | +/-4 | 51,295 | Anthropic | Proprietary |
| 18 | gpt-5.4 | 1,467 | +/-6 | 14,170 | OpenAI | Proprietary |
| 19 | qwen3.5-max-preview | 1,466 | +/-6 | 11,416 | Alibaba | Proprietary |
| 20 | deepseek-v4-pro | 1,463 | +/-9 | 4,163 | DeepSeek | MIT |
| 21 | Claude Sonnet 4.6 | 1,463 | +/-6 | 13,434 | Anthropic | Proprietary |
| 22 | Gemini 3.0 Flash (thinking-minimal) | 1,463 | +/-4 | 37,596 | Google Deep Mind | Proprietary |
| 23 | deepseek-v4-pro-thinking | 1,462 | +/-9 | 3,783 | DeepSeek | MIT |
| 24 | Grok 4.1 | 1,461 | +/-4 | 55,513 | xAI | Proprietary |
| 25 | dola-seed-2.0-pro | 1,460 | +/-5 | 22,902 | Bytedance | Proprietary |
| 26 | kimi-k2.6 | 1,458 | +/-9 | 4,355 | Moonshot | Modified MIT |
| 27 | gpt-5.4-mini-high | 1,457 | +/-6 | 11,237 | OpenAI | Proprietary |
| 28 | GLM-5 | 1,457 | +/-5 | 17,733 | 智谱AI | MIT |
| 29 | GPT-5.1 Pro | 1,455 | +/-4 | 40,871 | OpenAI | Proprietary |
| 30 | Claude Sonnet 4.5 (thinking-32k) | 1,453 | +/-3 | 63,725 | Anthropic | Proprietary |
| 31 | Claude Sonnet 4.5 | 1,452 | +/-3 | 61,608 | Anthropic | Proprietary |
| 32 | gemma-4-31b | 1,451 | +/-8 | 5,818 | Apache 2.0 | |
| 33 | ERNIE 5.0 | 1,450 | +/-4 | 25,815 | 百度 | Proprietary |
| 34 | gpt-5.3-chat-latest | 1,450 | +/-5 | 18,707 | OpenAI | Proprietary |
| 35 | ERNIE 5.0 | 1,449 | +/-7 | 9,767 | 百度 | Proprietary |
| 36 | Kimi K2 Thinking | 1,449 | +/-4 | 24,213 | Moonshot AI | Modified MIT |
| 37 | Opus 4.1 (thinking-16k) | 1,449 | +/-3 | 49,864 | Anthropic | Proprietary |
| 38 | Gemini 2.5 Pro Experimental 03-25 | 1,448 | +/-3 | 111,209 | Google Deep Mind | Proprietary |
| 39 | mimo-v2-pro | 1,448 | +/-6 | 12,059 | Xiaomi | Proprietary |
| 40 | Opus 4.1 | 1,447 | +/-3 | 77,453 | Anthropic | Proprietary |
| 41 | qwen3.6-plus | 1,447 | +/-8 | 5,480 | Alibaba | Proprietary |
| 42 | Qwen3.5-397B-A17B | 1,446 | +/-5 | 19,254 | 阿里巴巴 | Apache 2.0 |
| 43 | GPT-4.5 | 1,444 | +/-6 | 14,547 | OpenAI | Proprietary |
| 44 | chatgpt-4o-latest-20250326 | 1,443 | +/-3 | 82,567 | OpenAI | Proprietary |
| 45 | GLM-4.7 | 1,443 | +/-6 | 12,138 | 智谱AI | MIT |
| 46 | GPT-5.2 Pro | 1,440 | +/-4 | 34,410 | OpenAI | Proprietary |
| 47 | deepseek-v4-flash-thinking | 1,439 | +/-9 | 3,607 | DeepSeek | MIT |
| 48 | gemini-3.1-flash-lite-preview | 1,439 | +/-5 | 20,088 | Proprietary | |
| 49 | gemma-4-26b-a4b | 1,439 | +/-8 | 5,778 | Apache 2.0 | |
| 50 | GPT-5.1 Instant | 1,439 | +/-4 | 43,516 | OpenAI | Proprietary |
| 51 | GPT-5.2 | 1,439 | +/-4 | 31,577 | OpenAI | Proprietary |
| 52 | Qwen3 Max (Preview) | 1,435 | +/-5 | 27,769 | 阿里巴巴 | Proprietary |
| 53 | longcat-flash-chat-2602-exp | 1,434 | +/-6 | 9,660 | Meituan | Proprietary |
| 54 | GPT-5-Pro | 1,433 | +/-5 | 31,986 | OpenAI | Proprietary |
| 55 | deepseek-v4-flash | 1,433 | +/-9 | 3,497 | DeepSeek | MIT |
| 56 | kimi-k2.5-instant | 1,432 | +/-7 | 8,205 | Moonshot | Modified MIT |
| 57 | grok-4-1-fast-reasoning | 1,432 | +/-4 | 46,539 | xAI | Proprietary |
| 58 | OpenAI o3 | 1,431 | +/-4 | 59,795 | OpenAI | Proprietary |
| 59 | kimi-k2-thinking-turbo | 1,430 | +/-3 | 49,631 | Moonshot | Modified MIT |
| 60 | amazon-nova-experimental-chat-26-02-10 | 1,428 | +/-10 | 3,421 | Amazon | Proprietary |
| 61 | GPT-5 | 1,426 | +/-4 | 31,630 | OpenAI | Proprietary |
| 62 | GLM-4.6 | 1,426 | +/-4 | 35,711 | 智谱AI | MIT |
| 63 | DeepSeek V3.2-Exp (thinking) | 1,425 | +/-7 | 9,078 | DeepSeek-AI | MIT |
| 64 | DeepSeek V3.2 | 1,424 | +/-4 | 44,738 | DeepSeek-AI | MIT |
| 65 | qwen3-max-2025-09-23 | 1,424 | +/-6 | 9,181 | Alibaba | Proprietary |
| 66 | Claude Opus 4 (thinking-16k) | 1,424 | +/-4 | 36,951 | Anthropic | Proprietary |
| 67 | DeepSeek V3.2-Exp | 1,423 | +/-6 | 11,949 | DeepSeek-AI | MIT |
| 68 | Qwen3-235B-A22B-2507 | 1,423 | +/-3 | 85,263 | 阿里巴巴 | Apache 2.0 |
| 69 | DeepSeek V3.2 (thinking) | 1,422 | +/-4 | 38,982 | DeepSeek-AI | MIT |
| 70 | DeepSeek-R1-0528 | 1,422 | +/-6 | 18,474 | DeepSeek-AI | MIT |
| 71 | Grok 4 Fast | 1,421 | +/-8 | 6,826 | xAI | Proprietary |
| 72 | ERNIE 5.0 | 1,419 | +/-9 | 4,732 | 百度 | Proprietary |
| 73 | qwen3.5-122b-a10b | 1,418 | +/-5 | 16,040 | Alibaba | Apache 2.0 |
| 74 | kimi-k2-0905-preview | 1,418 | +/-7 | 11,800 | Moonshot | Modified MIT |
| 75 | DeepSeek-V3.1 | 1,418 | +/-6 | 14,992 | DeepSeek-AI | MIT |
| 76 | Kimi K2 | 1,417 | +/-5 | 27,668 | Moonshot AI | Modified MIT |
| 77 | deepseek-v3.1-terminus-thinking | 1,417 | +/-10 | 3,469 | DeepSeek | MIT |
| 78 | DeepSeek-V3.1 (thinking) | 1,417 | +/-7 | 11,759 | DeepSeek-AI | MIT |
| 79 | DeepSeek-V3.1 Terminus | 1,416 | +/-10 | 3,713 | DeepSeek-AI | MIT |
| 80 | amazon-nova-experimental-chat-26-01-10 | 1,416 | +/-10 | 3,420 | Amazon | Proprietary |
| 81 | Qwen3-VL-235B-A22B-Instruct | 1,416 | +/-6 | 11,528 | 阿里巴巴 | Apache 2.0 |
| 82 | Mistral Large 3 | 1,415 | +/-4 | 41,361 | MistralAI | Apache 2.0 |
| 83 | gpt-4.1-2025-04-14 | 1,413 | +/-4 | 51,065 | OpenAI | Proprietary |
| 84 | Claude Opus 4 | 1,412 | +/-4 | 44,273 | Anthropic | Proprietary |
| 85 | Grok 3 | 1,412 | +/-4 | 32,920 | xAI | Proprietary |
| 86 | GLM-4.5 | 1,411 | +/-5 | 24,352 | 智谱AI | MIT |
| 87 | Gemini 2.5 Flash | 1,411 | +/-3 | 110,831 | Google Deep Mind | Proprietary |
| 88 | grok-4-0709 | 1,410 | +/-4 | 41,428 | xAI | Proprietary |
| 89 | Magistral-Medium-2506 | 1,409 | +/-3 | 80,898 | MistralAI | Proprietary |
| 90 | claude-haiku-4-5-20251001 | 1,408 | +/-3 | 63,329 | Anthropic | Proprietary |
| 91 | qwen3.5-27b | 1,405 | +/-5 | 15,682 | Alibaba | Apache 2.0 |
| 92 | gemini-2.5-flash-preview-09-2025 | 1,405 | +/-4 | 32,953 | Proprietary | |
| 93 | gpt-5.4-nano-high | 1,405 | +/-6 | 10,607 | OpenAI | Proprietary |
| 94 | grok-4-fast-reasoning | 1,404 | +/-5 | 18,745 | xAI | Proprietary |
| 95 | MiniMax-M2.7 | 1,403 | +/-6 | 10,307 | MiniMaxAI | Modified MIT |
| 96 | qwen3-235b-a22b-no-thinking | 1,403 | +/-4 | 38,253 | Alibaba | Apache 2.0 |
| 97 | o1-2024-12-17 | 1,402 | +/-4 | 27,807 | OpenAI | Proprietary |
| 98 | qwen3-next-80b-a3b-instruct | 1,402 | +/-5 | 22,907 | Alibaba | Apache 2.0 |
| 99 | longcat-flash-chat | 1,401 | +/-6 | 11,417 | Meituan | MIT |
| 100 | MiniMax M2.5 | 1,399 | +/-5 | 21,236 | MiniMaxAI | Modified MIT |
| 101 | qwen3-235b-a22b-thinking-2507 | 1,399 | +/-7 | 9,008 | Alibaba | Apache 2.0 |
| 102 | Claude Sonnet 4 (thinking-32k) | 1,399 | +/-4 | 35,153 | Anthropic | Proprietary |
| 103 | DeepSeek-R1 | 1,398 | +/-5 | 18,524 | DeepSeek-AI | MIT |
| 104 | qwen3.5-flash | 1,397 | +/-5 | 16,394 | Alibaba | Proprietary |
| 105 | hunyuan-vision-1.5-thinking | 1,396 | +/-12 | 2,218 | Tencent | Proprietary |
| 106 | qwen3-vl-235b-a22b-thinking | 1,396 | +/-7 | 7,947 | Alibaba | Apache 2.0 |
| 107 | qwen3.5-35b-a3b | 1,395 | +/-5 | 16,250 | Alibaba | Apache 2.0 |
| 108 | amazon-nova-experimental-chat-12-10 | 1,395 | +/-10 | 3,687 | Amazon | Proprietary |
| 109 | DeepSeek-V3-0324 | 1,395 | +/-4 | 45,546 | DeepSeek-AI | MIT |
| 110 | mimo-v2-flash (non-thinking) | 1,393 | +/-4 | 33,906 | Xiaomi | MIT |
| 111 | mai-1-preview | 1,392 | +/-5 | 17,901 | Microsoft AI | Proprietary |
| 112 | Step 3.5 Flash | 1,392 | +/-5 | 21,987 | StepFunAI | Apache 2.0 |
| 113 | gpt-5-mini-high | 1,390 | +/-5 | 27,067 | OpenAI | Proprietary |
| 114 | o4-mini-2025-04-16 | 1,390 | +/-4 | 45,482 | OpenAI | Proprietary |
| 115 | Claude Sonnet 4 | 1,389 | +/-4 | 40,359 | Anthropic | Proprietary |
| 116 | o1-preview | 1,388 | +/-5 | 31,122 | OpenAI | Proprietary |
| 117 | hunyuan-t1-20250711 | 1,387 | +/-9 | 4,714 | Tencent | Proprietary |
| 118 | mimo-v2-flash (thinking) | 1,387 | +/-6 | 10,980 | Xiaomi | MIT |
| 119 | qwen3-coder-480b-a35b-instruct | 1,387 | +/-5 | 25,762 | Alibaba | Apache 2.0 |
| 120 | Claude Sonnet 3.7 (thinking-32k) | 1,387 | +/-4 | 38,841 | Anthropic | Proprietary |
| 121 | mistral-medium-2505 | 1,387 | +/-5 | 33,260 | Mistral | Proprietary |
| 122 | minimax-m2.1-preview | 1,385 | +/-5 | 17,150 | MiniMax | MIT |
| 123 | qwen3-30b-a3b-instruct-2507 | 1,383 | +/-5 | 23,768 | Alibaba | Apache 2.0 |
| 124 | hunyuan-turbos-20250416 | 1,382 | +/-6 | 10,723 | Tencent | Proprietary |
| 125 | gpt-4.1-mini-2025-04-14 | 1,382 | +/-4 | 39,371 | OpenAI | Proprietary |
| 126 | gemini-2.5-flash-lite-preview-09-2025-no-thinking | 1,380 | +/-3 | 47,291 | Proprietary | |
| 127 | GLM-4.6V | 1,378 | +/-11 | 2,806 | 智谱AI | MIT |
| 128 | qwen3-235b-a22b | 1,375 | +/-5 | 26,285 | Alibaba | Apache 2.0 |
| 129 | trinity-large-preview | 1,375 | +/-5 | 17,170 | Arcee AI | Apache 2.0 |
| 130 | gemini-2.5-flash-lite-preview-06-17-thinking | 1,374 | +/-5 | 32,957 | Proprietary | |
| 131 | qwen2.5-max | 1,374 | +/-4 | 32,629 | Alibaba | Proprietary |
| 132 | glm-4.5-air | 1,373 | +/-4 | 31,136 | Z.ai | MIT |
| 133 | claude-3-5-sonnet-20241022 | 1,372 | +/-3 | 88,367 | Anthropic | Proprietary |
| 134 | Claude Sonnet 3.7 | 1,371 | +/-4 | 43,216 | Anthropic | Proprietary |
| 135 | qwen3-next-80b-a3b-thinking | 1,369 | +/-6 | 13,718 | Alibaba | Apache 2.0 |
| 136 | glm-4.7-flash | 1,368 | +/-6 | 11,769 | Z.ai | MIT |
| 137 | amazon-nova-experimental-chat-11-10 | 1,367 | +/-4 | 25,430 | Amazon | Proprietary |
| 138 | gemma-3-27b-it | 1,366 | +/-4 | 47,577 | Gemma | |
| 139 | minimax-m1 | 1,363 | +/-4 | 35,252 | MiniMax | Apache 2.0 |
| 140 | o3-mini-high | 1,363 | +/-5 | 18,589 | OpenAI | Proprietary |
| 141 | grok-3-mini-high | 1,362 | +/-5 | 16,979 | xAI | Proprietary |
| 142 | nvidia-nemotron-3-super-120b-a12b | 1,361 | +/-7 | 7,408 | Nvidia | NVIDIA Open Model |
| 143 | gemini-2.0-flash-001 | 1,360 | +/-4 | 43,771 | Proprietary | |
| 144 | deepseek-v3 | 1,358 | +/-5 | 21,770 | DeepSeek | DeepSeek |
| 145 | mistral-small-2506 | 1,357 | +/-5 | 17,722 | Mistral | Apache 2.0 |
| 146 | grok-3-mini-beta | 1,357 | +/-5 | 22,722 | xAI | Proprietary |
| 147 | intellect-3 | 1,356 | +/-8 | 5,332 | Prime Intellect | MIT |
| 148 | command-a-03-2025 | 1,354 | +/-3 | 56,346 | Cohere | CC-BY-NC-4.0 |
| 149 | gpt-oss-120b | 1,353 | +/-4 | 30,674 | OpenAI | Apache 2.0 |
| 150 | glm-4.5v | 1,353 | +/-8 | 4,968 | Z.ai | MIT |
| 151 | gemini-2.0-flash-lite-preview-02-05 | 1,353 | +/-4 | 24,955 | Proprietary | |
| 152 | gemini-1.5-pro-002 | 1,351 | +/-3 | 55,606 | Proprietary | |
| 153 | amazon-nova-experimental-chat-10-20 | 1,350 | +/-6 | 11,486 | Amazon | Proprietary |
| 154 | hunyuan-turbos-20250226 | 1,349 | +/-12 | 2,220 | Tencent | Proprietary |
| 155 | step-3 | 1,348 | +/-7 | 6,560 | StepFun | Apache 2.0 |
| 156 | o3-mini | 1,348 | +/-4 | 57,375 | OpenAI | Proprietary |
| 157 | amazon-nova-experimental-chat-10-09 | 1,347 | +/-11 | 2,839 | Amazon | Proprietary |
| 158 | qwen3-32b | 1,347 | +/-9 | 3,926 | Alibaba | Apache 2.0 |
| 159 | llama-3.1-nemotron-ultra-253b-v1 | 1,347 | +/-12 | 2,549 | Nvidia | Nvidia Open Model |
| 160 | mercury-2 | 1,347 | +/-11 | 3,124 | Inception AI | Proprietary |
| 161 | minimax-m2 | 1,346 | +/-8 | 6,876 | MiniMax | Apache 2.0 |
| 162 | qwen-plus-0125 | 1,346 | +/-8 | 5,819 | Alibaba | Proprietary |
| 163 | ling-flash-2.0 | 1,346 | +/-7 | 7,018 | InclusionAI | MIT |
| 164 | gpt-4o-2024-05-13 | 1,345 | +/-3 | 112,881 | OpenAI | Proprietary |
| 165 | glm-4-plus-0111 | 1,343 | +/-8 | 5,760 | Zhipu | Proprietary |
| 166 | nvidia-llama-3.3-nemotron-super-49b-v1.5 | 1,343 | +/-10 | 3,347 | Nvidia | Nvidia Open |
| 167 | claude-3-5-sonnet-20240620 | 1,342 | +/-3 | 82,419 | Anthropic | Proprietary |
| 168 | gemma-3-12b-it | 1,342 | +/-10 | 3,829 | Gemma | |
| 169 | hunyuan-turbo-0110 | 1,340 | +/-12 | 2,290 | Tencent | Proprietary |
| 170 | nova-2-lite | 1,337 | +/-6 | 12,251 | Amazon | Proprietary |
| 171 | gpt-5-nano-high | 1,337 | +/-7 | 8,281 | OpenAI | Proprietary |
| 172 | o1-mini | 1,337 | +/-4 | 51,981 | OpenAI | Proprietary |
| 173 | qwq-32b | 1,336 | +/-4 | 25,406 | Alibaba | Apache 2.0 |
| 174 | grok-2-2024-08-13 | 1,335 | +/-4 | 63,498 | xAI | Proprietary |
| 175 | gemini-advanced-0514 | 1,335 | +/-5 | 50,148 | Proprietary | |
| 176 | gpt-4o-2024-08-06 | 1,335 | +/-4 | 45,499 | OpenAI | Proprietary |
| 177 | llama-3.1-405b-instruct-bf16 | 1,335 | +/-4 | 41,375 | Meta | Llama 3.1 Community |
| 178 | step-2-16k-exp-202412 | 1,334 | +/-9 | 4,833 | StepFun | Proprietary |
| 179 | llama-3.1-405b-instruct-fp8 | 1,333 | +/-4 | 59,656 | Meta | Llama 3.1 Community |
| 180 | olmo-3.1-32b-instruct | 1,331 | +/-6 | 12,238 | Ai2 | Apache 2.0 |
| 181 | yi-lightning | 1,328 | +/-5 | 27,332 | 01 AI | Proprietary |
| 182 | molmo-2-8b | 1,328 | +/-21 | 803 | Ai2 | Apache 2.0 |
| 183 | llama-3.3-nemotron-49b-super-v1 | 1,327 | +/-12 | 2,218 | Nvidia | Nvidia |
| 184 | qwen3-30b-a3b | 1,327 | +/-5 | 26,507 | Alibaba | Apache 2.0 |
| 185 | llama-4-maverick-17b-128e-instruct | 1,327 | +/-4 | 40,010 | Meta | Llama 4 |
| 186 | hunyuan-large-2025-02-10 | 1,326 | +/-10 | 3,738 | Tencent | Proprietary |
| 187 | gpt-4-turbo-2024-04-09 | 1,324 | +/-4 | 98,114 | OpenAI | Proprietary |
| 188 | deepseek-v2.5-1210 | 1,323 | +/-8 | 6,795 | DeepSeek | DeepSeek |
| 189 | gemini-1.5-pro-001 | 1,323 | +/-4 | 79,138 | Proprietary | |
| 190 | claude-3-5-haiku-20241022 | 1,323 | +/-3 | 70,029 | Anthropic | Proprietary |
| 191 | llama-4-scout-17b-16e-instruct | 1,322 | +/-5 | 30,319 | Meta | Llama |
| 192 | gpt-4.1-nano-2025-04-14 | 1,322 | +/-8 | 6,103 | OpenAI | Proprietary |
| 193 | Claude3-Opus | 1,321 | +/-3 | 194,909 | Anthropic | Proprietary |
| 194 | ring-flash-2.0 | 1,321 | +/-7 | 7,158 | InclusionAI | MIT |
| 195 | step-1o-turbo-202506 | 1,320 | +/-7 | 9,044 | StepFun | Proprietary |
| 196 | glm-4-plus | 1,319 | +/-5 | 26,126 | Zhipu AI | Proprietary |
| 197 | llama-3.3-70b-instruct | 1,318 | +/-3 | 54,756 | Meta | Llama-3.3 |
| 198 | gemma-3n-e4b-it | 1,318 | +/-5 | 22,616 | Gemma | |
| 199 | qwen-max-0919 | 1,318 | +/-6 | 16,478 | Alibaba | Qwen |
| 200 | gpt-oss-20b | 1,317 | +/-6 | 10,640 | OpenAI | Apache 2.0 |
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)
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。
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——击败强模型加分更多，输给弱模型扣分也更多。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。
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。
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（国产模型如 Qwen、GLM、DeepSeek 在中文场景更占优）、开源程度（MIT/Apache 协议可商用）以及响应速度（Flash/mini 轻量版时延更低）。