跳到正文

大模型排行榜

Arena

比较模型针对图像给出的回答,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 (High)anthropic1,3251318–133215,449
2Claude Fable 5.1 (Max)anthropic1,3221314–13307,185
3Claude Opus 5.5 (High)anthropic1,3221313–13315,903
4Claude Opus 4.7anthropic1,3171311–132424,930
5Claude Opus 5 (High)anthropic1,3171311–132419,229
6Gemini 3.8 Flash (High)google1,3161309–13249,782
7Claude Opus 4.6 (High)anthropic1,3151309–132122,067
8Claude Opus 4.7 (High)anthropic1,3141307–132022,530
9Gemini 3.7 Flash (High)google1,3141305–13226,776
10Claude Fable 5.2 (Max)anthropic1,3131295–13311,176
11Claude Opus 4.6anthropic1,3121306–131826,552
12Gemini 3.5 Flash (High)google1,3081301–131415,228
13Qwen3.8-Maxalibaba1,3071299–131410,833
14Muse Spark 1.3 (Max)meta1,3061299–13149,287
15Gemini 3.5 Flash (Medium)google1,3061300–131315,320
16Muse Sparkmeta1,3061297–13155,835
17Gemini 3 Progoogle1,3051297–131213,482
18Muse Spark 1.2 (xHigh)meta1,3041290–13191,922
19GPT-5.4 (High)openai1,3011295–130631,907
20Muse Spark 1.3 (xHigh)meta1,3001293–13089,809
21Gemini 3.6 Flash (High)google1,2971289–130410,275
22Gemini 3.1 Pro Previewgoogle1,2961292–130148,038
23GLM-5.3-Flashzai1,2961291–130236,497
24GPT-5.5openai1,2961290–130229,925
25GPT-5.5 (High)openai1,2951290–130128,221
26Claude Opus 4.8 (High)anthropic1,2941288–130118,113
27GPT-5.4openai1,2911285–129822,546
28Muse Spark 1.1meta1,2911284–129813,311
29Claude Opus 4.8anthropic1,2881282–129418,588
30Grok 4.5xai1,2881281–129513,903
31Claude Sonnet 5.5 (xHigh)anthropic1,2861277–12955,388
32Gemini 3 Flashgoogle1,2851280–129038,733
33GPT-6.1 Sol (Max)openai1,2841274–12935,043
34Claude Sonnet 4.6anthropic1,2821276–128927,131
35Kimi K2.6moonshot1,2821276–128916,241
36GPT-5.6 Sol (xHigh)openai1,2821275–128912,244
37Gemma 4 31Bgoogle1,2791274–128542,217
38DeepSeek-V4.1-Flash-20260910 (Max)deepseek1,2781268–12884,192
39Qwen3.7-Plusalibaba1,2771271–128415,084
40GPT-6 Astra (Max)openai1,2761268–12856,436
41Dola Seed 2.0 Probytedance1,2741267–128110,914
42Claude Sonnet 5 (High)anthropic1,2721265–127814,701
43Qwen3.8-27Balibaba1,2711264–127810,910
44GPT-5.6 Terra (xHigh)openai1,2701263–127713,166
45Kimi K2.5 (Thinking)moonshot1,2691264–127527,947
46Gemini 3.5 Flash-Litegoogle1,2681261–127510,713
47GPT-5.2 Chat (20260210)openai1,2681261–127416,336
48MiMo-V2.6-Proxiaomi1,2661251–12811,812
49Gemini 3 Flash (Minimal)google1,2661260–127136,810
50Grok 4.6 (High)xai1,2651257–12738,578

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)