跳到正文

大模型排行榜

Arena

比较模型生成网页的效果,由参与者投票。

Arena AI 原榜 · 榜单日期

排名模型评分对战票数
1Claude Fable 5.2 (xHigh)anthropic1,8231807–18391,993
2Claude Opus 5.5 (Max)anthropic1,8181807–18294,912
3Claude Sonnet 5.5 (xHigh)anthropic1,7711762–17816,721
4Gemini 4 Carbon (High)google1,7691753–17842,092
5Gemini 4 Neon (High)google1,7631749–17772,324
6GPT-6 Astra (Max)openai1,7631755–17719,316
7GPT-6.2 Astra (Max)openai1,7591744–17751,848
8GPT-6.2 Sol (Max)openai1,7551740–17691,947
9Qwen4-Max (xHigh)alibaba1,7501734–17651,808
10GPT-6.1 Sol (Max)openai1,7441734–17536,459
11Claude Fable 5.1 (Max)anthropic1,7411734–174910,591
12Muse Wonder Full (Max)meta1,7231709–17382,052
13Claude Sonnet 5.5 (High)anthropic1,7161708–17248,851
14Gemini 4 Hydrogen (High)google1,7011687–17152,039
15Claude Opus 5 (Max)anthropic1,6921686–169819,465
16GPT-6.1 Terra (Max)openai1,6841675–16945,170
17Kimi K3.1 Pro (Max)moonshot1,6831674–16925,958
18Gemini 4 Argon (High)google1,6811672–16905,978
19Qwen3.8-Max (0902)alibaba1,6751669–168125,208
20GPT-6 Sol (Max)openai1,6741666–16828,303
21Qwen3.8-Maxalibaba1,6711659–16833,356
22DeepSeek-V4.1-Flash-20261017 (Max)deepseek1,6591652–166517,765
23Kimi K3 (Max)moonshot1,6581653–166423,328
24Claude Opus 5 (High)anthropic1,6571651–166226,790
25Muse Spark 1.3 (Max)meta1,6541647–166013,332
26Hy4tencent1,6491637–16612,924
27Muse Wonder (Max)meta1,6481639–16566,477
28Inkling 2 Smallthinky1,6441626–16611,195
29Grok 4.7 (xHigh)xai1,6381630–16459,273
30MiMo-V2.6-Flashxiaomi1,6381630–16459,966
31Hy4 previewtencent1,6321622–16414,577
32Qwen3.8-Flash-Nextalibaba1,6261619–163312,404
33Claude Fable 5 (High)anthropic1,6261620–163216,874
34GLM-5.3 (Max)zai1,6221615–162813,737
35MiMo-V2.6-Proxiaomi1,6221615–162910,177
36DeepSeek-V4.1-Flash-20260910 (Max)deepseek1,6211610–16323,377
37Muse Spark 1.3 (xHigh)meta1,6211614–162812,861
38Grok 4.6 (High)xai1,6181611–162512,121
39GPT-5.6 Sol (xHigh, Codex Harness)openai1,6151610–162118,792
40GLM-5.3-Flashzai1,6101604–161624,714
41Muse Ember (Max)meta1,6091592–16261,322
42U2 Flashunisound1,6091592–16251,499
43GLM-5.2 (Max)zai1,6071602–161317,837
44Claude Haiku 5.5 (High)anthropic1,5961588–16048,314
45Qwen3.8-27Balibaba1,5891583–159522,065
46Gemini 3.7 Flash (High)google1,5891582–159514,496
47Gemini 3.8 Flash (High)google1,5841577–159015,072
48DeepSeek-V4-Flash (High)deepseek1,5821572–15914,809
49MiniMax-M3.1-Flashminimax1,5821573–15915,948
50DeepSeek-V4-Pro (High, 20260813)deepseek1,5811571–15904,593

公开数据集 · CC BY 4.0

展示前 50 条,沿用原始名次。评分取整,下方为 95% 置信区间。

最近同步 (北京时间)