跳到正文

大模型排行榜

Arena

人在盲测中对文本回答作出偏好选择。

Arena AI 原榜 · 榜单日期

排名模型评分对战票数
1Gemini 4 Carbon (High)google1,5391530–15494,262
2Gemini 4 Argon (High)google1,5281523–153327,193
3Gemini 4 Neon (High)google1,5241515–15325,555
4Claude Opus 5.5 (High)anthropic1,5131507–151918,286
5Claude Fable 5.1 (Max)anthropic1,5121507–151822,408
6Claude Opus 5 (Max)anthropic1,5061502–151135,534
7Muse Wonder (Max)meta1,5031498–150916,158
8Claude Opus 5 (High)anthropic1,5021499–150673,414
9Gemini 4 Hydrogen (High)google1,5021493–15114,743
10Claude Opus 4.6 (High)anthropic1,5011498–150584,188
11Gemini 3.8 Flash (High)google1,5001496–150473,351
12Claude Opus 4.6anthropic1,4971494–150088,918
13Muse Spark 1.3 (Max)meta1,4941489–149830,142
14MiMo-V2.6-Proxiaomi1,4931483–15024,045
15Claude Fable 5 (High)anthropic1,4921488–149645,315
16Claude Opus 4.7 (High)anthropic1,4901486–149464,074
17Gemini 3.7 Flash (High)google1,4891485–149433,087
18Qwen3.8-Maxalibaba1,4851481–149040,248
19Muse Spark 1.2 (xHigh)meta1,4841479–149018,264
20Claude Opus 4.7anthropic1,4831480–148765,152
21Gemini 3.5 Flash (High)google1,4831479–148747,470
22Gemini 3.1 Pro Previewgoogle1,4801477–1483131,889
23Gemini 3 Progoogle1,4801476–148341,792
24Gemini 3.6 Flash (High)google1,4801475–148435,275
25Muse Spark 1.1meta1,4791475–148345,342
26Claude Fable 5.2 (Max)anthropic1,4781470–14875,216
27Gemini 3.5 Flash (Medium)google1,4781474–148246,163
28GLM-5.3 (Max)zai1,4751471–148034,708
29Kimi K3 (Max)moonshot1,4751471–147944,364
30Qwen3.7-Max-Previewalibaba1,4751465–14843,888
31Muse Sparkmeta1,4741469–148013,972
32GLM-5.3-Flashzai1,4711467–147656,089
33Qwen3.5-Max-Previewalibaba1,4711466–147622,361
34GPT-5.5 (High)openai1,4711467–147568,109
35Claude Sonnet 5.5 (xHigh)anthropic1,4701464–147516,925
36GLM-5.2 (Max)zai1,4691465–147453,978
37GPT-5.4 (High)openai1,4691465–147363,928
38ERNIE 5.1baidu1,4681463–147238,932
39GPT-5.5openai1,4661463–147069,700
40Gemini 3 Flashgoogle1,4661462–147031,095
41MiMo-V2.5-Proxiaomi1,4641460–146776,511
42DeepSeek-V4.1-Flash-20260910 (Max)deepseek1,4631457–146913,492
43GLM-5.1zai1,4631459–146666,799
44Claude Opus 4.8 (High)anthropic1,4621458–146562,925
45Gemini 2.5 Progoogle1,4581455–1460124,685
46Kimi K3.1 Pro (Max)moonshot1,4581450–14666,658
47Claude Sonnet 4.6anthropic1,4581454–146169,794
48GPT-5.6 Sol (xHigh)openai1,4571453–146243,782
49Kimi K2.6moonshot1,4551451–146039,440
50Qwen3.7-Plusalibaba1,4541450–145841,181

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)