AI 智能体模型排行榜

查看 Arena 智能体任务榜的净改善率。

评测来源 Arena · 智能体原榜发布 查看原榜评分方法

净改善率

52 个模型
Arena · 智能体 · 净改善率
排名模型 / 开发机构净改善率会话95% CI许可详情
01Claude Opus 5.5 (High)Anthropic14.33%净改善率会话610395% CI12.15% / 16.51%许可Proprietary会话610395% CI12.15% / 16.51%许可Proprietary来源
02GPT 6 Astra (Max)OpenAI13.09%净改善率会话1202995% CI10.73% / 15.46%许可Proprietary会话1202995% CI10.73% / 15.46%许可Proprietary详情
03Claude Fable 5.1 (Max)Anthropic12.66%净改善率会话1553495% CI10.61% / 14.72%许可Proprietary会话1553495% CI10.61% / 14.72%许可Proprietary详情
04Claude Sonnet 5.5 (Max)Anthropic11.95%净改善率会话678095% CI8.64% / 15.27%许可Proprietary会话678095% CI8.64% / 15.27%许可Proprietary来源
05GPT 6.1 Sol (Max)OpenAI11.72%净改善率会话820195% CI8.68% / 14.77%许可Proprietary会话820195% CI8.68% / 14.77%许可Proprietary来源
06GPT 6 Sol (Max)OpenAI9.92%净改善率会话610095% CI7.54% / 12.30%许可Proprietary会话610095% CI7.54% / 12.30%许可Proprietary来源
07Gemini 4 Argon (High)Google9.26%净改善率会话1796395% CI6.90% / 11.62%许可Proprietary会话1796395% CI6.90% / 11.62%许可Proprietary来源
08Claude Fable 5 (High)Anthropic8.94%净改善率会话4190895% CI7.67% / 10.21%许可Proprietary会话4190895% CI7.67% / 10.21%许可Proprietary详情
09Claude Opus 5 (Max)Anthropic8.12%净改善率会话2281895% CI6.64% / 9.59%许可Proprietary会话2281895% CI6.64% / 9.59%许可Proprietary详情
10Claude Opus 5 (High)Anthropic7.95%净改善率会话2755395% CI6.56% / 9.34%许可Proprietary会话2755395% CI6.56% / 9.34%许可Proprietary详情
11Claude Opus 4.8 (High)Anthropic7.13%净改善率会话4030895% CI5.76% / 8.50%许可Proprietary会话4030895% CI5.76% / 8.50%许可Proprietary详情
12GPT 5.6 Sol (xHigh)OpenAI6.23%净改善率会话3576895% CI4.95% / 7.52%许可Proprietary会话3576895% CI4.95% / 7.52%许可Proprietary详情
13Claude Sonnet 5 (High)Anthropic4.91%净改善率会话3237295% CI3.23% / 6.60%许可Proprietary会话3237295% CI3.23% / 6.60%许可Proprietary详情
14GPT 5.5 (xHigh)OpenAI4.38%净改善率会话5438595% CI3.28% / 5.47%许可Proprietary会话5438595% CI3.28% / 5.47%许可Proprietary详情
15Muse Spark 1.3 (Max)Meta4.04%净改善率会话7076695% CI3.38% / 4.71%许可Proprietary会话7076695% CI3.38% / 4.71%许可Proprietary详情
16Deepseek V4.1 Flash (Max)DeepSeek3.80%净改善率会话18269395% CI3.40% / 4.19%许可MIT会话18269395% CI3.40% / 4.19%许可MIT详情
17Kimi K3 (Max)Moonshot AI3.78%净改善率会话13321795% CI3.19% / 4.37%许可Kimi K3 license会话13321795% CI3.19% / 4.37%许可Kimi K3 license详情
18Hy4 previewTencent3.61%净改善率会话7070295% CI2.91% / 4.31%许可Apache 2.0会话7070295% CI2.91% / 4.31%许可Apache 2.0详情
19MiMo V2.6 ProXiaomi3.52%净改善率会话1137695% CI1.61% / 5.44%许可MIT会话1137695% CI1.61% / 5.44%许可MIT来源
20GLM 5.2 (Max)Z.ai3.19%净改善率会话8592495% CI2.46% / 3.92%许可MIT会话8592495% CI2.46% / 3.92%许可MIT详情
21Grok 4.7 (xHigh)xAI2.83%净改善率会话1321495% CI1.12% / 4.54%许可Proprietary会话1321495% CI1.12% / 4.54%许可Proprietary详情
22Gemini 3.8 Flash (High)Google2.75%净改善率会话3598395% CI1.82% / 3.68%许可Proprietary会话3598395% CI1.82% / 3.68%许可Proprietary详情
23Qwen3.8 MaxAlibaba2.35%净改善率会话5999595% CI1.75% / 2.94%许可Proprietary会话5999595% CI1.75% / 2.94%许可Proprietary详情
24GLM 5.3 (Max)Z.ai2.15%净改善率会话8580695% CI1.43% / 2.86%许可MIT会话8580695% CI1.43% / 2.86%许可MIT详情
25GPT 6 Luna (Max)OpenAI1.93%净改善率会话10909895% CI1.29% / 2.57%许可Proprietary会话10909895% CI1.29% / 2.57%许可Proprietary来源
26DeepSeek V4 Pro (High) (0813)DeepSeek1.34%净改善率会话708195% CI-0.32% / 3.00%许可MIT会话708195% CI-0.32% / 3.00%许可MIT详情
27Grok 4.5xAI1.19%净改善率会话4188295% CI0.17% / 2.22%许可Proprietary会话4188295% CI0.17% / 2.22%许可Proprietary详情
28GPT 5.4 (High)OpenAI1.19%净改善率会话8189895% CI0.23% / 2.14%许可Proprietary会话8189895% CI0.23% / 2.14%许可Proprietary详情
29Grok 4.6 (xHigh)xAI1.05%净改善率会话2799095% CI0.05% / 2.05%许可Proprietary会话2799095% CI0.05% / 2.05%许可Proprietary详情
30GPT 5.5OpenAI0.93%净改善率会话8267195% CI-0.08% / 1.94%许可Proprietary会话8267195% CI-0.08% / 1.94%许可Proprietary详情
31GPT 5.6 Terra (xHigh)OpenAI0.88%净改善率会话2332695% CI-0.32% / 2.08%许可Proprietary会话2332695% CI-0.32% / 2.08%许可Proprietary详情
32MiMo V2.6 FlashXiaomi0.35%净改善率会话1686495% CI-1.12% / 1.83%许可MIT会话1686495% CI-1.12% / 1.83%许可MIT来源
33Step 5 PreviewStepFun0.01%净改善率会话1147395% CI-1.57% / 1.58%许可Proprietary会话1147395% CI-1.57% / 1.58%许可Proprietary来源
34Qwen3.8 Flash NextAlibaba-0.72%净改善率会话8994895% CI-1.34% / -0.10%许可qwen-community-1.0会话8994895% CI-1.34% / -0.10%许可qwen-community-1.0详情
35GLM 5.3 FlashZ.ai-0.97%净改善率会话8231495% CI-1.50% / -0.44%许可MIT会话8231495% CI-1.50% / -0.44%许可MIT详情
36GPT 5.6 Luna (xHigh)OpenAI-1.48%净改善率会话4326495% CI-2.17% / -0.79%许可Proprietary会话4326495% CI-2.17% / -0.79%许可Proprietary详情
37Qwen 3.8 27BAlibaba-2.01%净改善率会话5866295% CI-2.64% / -1.38%许可Apache 2.0会话5866295% CI-2.64% / -1.38%许可Apache 2.0详情
38Gemini 3.7 Flash (High)Google-2.10%净改善率会话5435895% CI-2.90% / -1.31%许可Proprietary会话5435895% CI-2.90% / -1.31%许可Proprietary详情
39Muse Spark 1.2 (xHigh)Meta-3.65%净改善率会话4971295% CI-4.37% / -2.94%许可Proprietary会话4971295% CI-4.37% / -2.94%许可Proprietary详情
40Muse Spark 1.1Meta-5.00%净改善率会话14422795% CI-5.51% / -4.49%许可Proprietary会话14422795% CI-5.51% / -4.49%许可Proprietary详情
41Hy3Tencent-5.06%净改善率会话3040295% CI-6.29% / -3.83%许可Apache 2.0会话3040295% CI-6.29% / -3.83%许可Apache 2.0详情
42Qwen3.7 MaxAlibaba-5.50%净改善率会话4113995% CI-6.46% / -4.54%许可Proprietary会话4113995% CI-6.46% / -4.54%许可Proprietary详情
43mistral-large-4Mistral AI-6.56%净改善率会话507895% CI-9.51% / -3.61%许可Proprietary会话507895% CI-9.51% / -3.61%许可Proprietary来源
44Minimax M3MiniMax-6.95%净改善率会话4204295% CI-7.86% / -6.05%许可MiniMax Community License会话4204295% CI-7.86% / -6.05%许可MiniMax Community License详情
45Qwen3.7 PlusAlibaba-7.21%净改善率会话2248395% CI-8.67% / -5.76%许可Proprietary会话2248395% CI-8.67% / -5.76%许可Proprietary详情
46Mimo V2.5 ProXiaomi-7.78%净改善率会话4671595% CI-8.68% / -6.89%许可MIT会话4671595% CI-8.68% / -6.89%许可MIT详情
47Gemini 3.6 Flash (High)Google-7.79%净改善率会话2288195% CI-8.85% / -6.73%许可Proprietary会话2288195% CI-8.85% / -6.73%许可Proprietary详情
48Gemini 3.1 Pro PreviewGoogle-7.79%净改善率会话8679495% CI-8.79% / -6.79%许可Proprietary会话8679495% CI-8.79% / -6.79%许可Proprietary详情
49Inkling Smallthinky-10.32%净改善率会话1356195% CI-11.76% / -8.88%许可Apache 2.0会话1356195% CI-11.76% / -8.88%许可Apache 2.0详情
50Inklingthinky-11.51%净改善率会话4308495% CI-12.63% / -10.39%许可Apache 2.0会话4308495% CI-12.63% / -10.39%许可Apache 2.0详情
显示更多已显示 50 / 52

数据来自 Arena 官方公开数据集(CC BY 4.0);保留原榜名次与评分。 官方数据集

未公布的数据记为 —