AI 智能体模型排行榜
查看 Arena 智能体任务榜的净改善率。
| 排名 | 模型 / 开发机构 | 净改善率 | 会话 | 95% CI | 许可 | 详情 | |
|---|---|---|---|---|---|---|---|
| 01 | Claude Opus 5.5 (Max)Anthropic | 20.14%净改善率 | 会话14169 | 95% CI18.06% / 22.23% | 许可Proprietary | 会话1416995% CI18.06% / 22.23%许可Proprietary | 来源 |
| 02 | Gemini 4 Carbon (High)Google | 16.72%净改善率 | 会话44873 | 95% CI15.95% / 17.48% | 许可Proprietary | 会话4487395% CI15.95% / 17.48%许可Proprietary | 来源 |
| 03 | Claude Sonnet 5.5 (Max)Anthropic | 15.42%净改善率 | 会话8596 | 95% CI12.65% / 18.20% | 许可Proprietary | 会话859695% CI12.65% / 18.20%许可Proprietary | 来源 |
| 04 | GPT-6 Astra (Max)OpenAI | 14.70%净改善率 | 会话15475 | 95% CI12.63% / 16.76% | 许可Proprietary | 会话1547595% CI12.63% / 16.76%许可Proprietary | 来源 |
| 05 | Claude Fable 5.1 (Max)Anthropic | 13.86%净改善率 | 会话17518 | 95% CI11.98% / 15.73% | 许可Proprietary | 会话1751895% CI11.98% / 15.73%许可Proprietary | 详情 |
| 06 | GPT-6.1 Sol (Max)OpenAI | 12.51%净改善率 | 会话30285 | 95% CI11.41% / 13.62% | 许可Proprietary | 会话3028595% CI11.41% / 13.62%许可Proprietary | 来源 |
| 07 | GPT-6.1 Terra (Max)OpenAI | 11.40%净改善率 | 会话21355 | 95% CI10.00% / 12.79% | 许可Proprietary | 会话2135595% CI10.00% / 12.79%许可Proprietary | 来源 |
| 08 | Gemini 4 Argon (High)Google | 10.53%净改善率 | 会话52818 | 95% CI9.68% / 11.37% | 许可Proprietary | 会话5281895% CI9.68% / 11.37%许可Proprietary | 来源 |
| 09 | Muse Wonder Full (Max)Meta | 9.42%净改善率 | 会话32592 | 95% CI8.48% / 10.37% | 许可Proprietary | 会话3259295% CI8.48% / 10.37%许可Proprietary | 来源 |
| 10 | Claude Fable 5 (High)Anthropic | 7.82%净改善率 | 会话51251 | 95% CI6.71% / 8.94% | 许可Proprietary | 会话5125195% CI6.71% / 8.94%许可Proprietary | 详情 |
| 11 | Claude Opus 5 (Max)Anthropic | 7.66%净改善率 | 会话32048 | 95% CI6.42% / 8.91% | 许可Proprietary | 会话3204895% CI6.42% / 8.91%许可Proprietary | 详情 |
| 12 | GPT-6 Sol (Max)OpenAI | 6.78%净改善率 | 会话12670 | 95% CI5.23% / 8.32% | 许可Proprietary | 会话1267095% CI5.23% / 8.32%许可Proprietary | 来源 |
| 13 | Hy4Tencent | 5.39%净改善率 | 会话12700 | 95% CI3.68% / 7.10% | 许可Proprietary | 会话1270095% CI3.68% / 7.10%许可Proprietary | 来源 |
| 14 | Kimi K3.1 Pro (Max)Moonshot AI | 5.29%净改善率 | 会话18045 | 95% CI4.01% / 6.57% | 许可Kimi K3.1 Pro License | 会话1804595% CI4.01% / 6.57%许可Kimi K3.1 Pro License | 来源 |
| 15 | GPT-5.6 Sol (xHigh)OpenAI | 4.57%净改善率 | 会话45533 | 95% CI3.54% / 5.61% | 许可Proprietary | 会话4553395% CI3.54% / 5.61%许可Proprietary | 来源 |
| 16 | MiMo-V2.6-ProXiaomi | 2.68%净改善率 | 会话48735 | 95% CI1.77% / 3.60% | 许可MIT | 会话4873595% CI1.77% / 3.60%许可MIT | 来源 |
| 17 | Grok 4.7 (xHigh)xAI | 2.68%净改善率 | 会话18917 | 95% CI1.22% / 4.14% | 许可Proprietary | 会话1891795% CI1.22% / 4.14%许可Proprietary | 详情 |
| 18 | Gemini 3.8 Flash (High)Google | 2.58%净改善率 | 会话53068 | 95% CI1.75% / 3.41% | 许可Proprietary | 会话5306895% CI1.75% / 3.41%许可Proprietary | 详情 |
| 19 | Muse Spark 1.3 (Max)Meta | 2.51%净改善率 | 会话81053 | 95% CI1.68% / 3.34% | 许可Proprietary | 会话8105395% CI1.68% / 3.34%许可Proprietary | 详情 |
| 20 | Claude Sonnet 5 (High)Anthropic | 2.44%净改善率 | 会话41857 | 95% CI1.28% / 3.61% | 许可Proprietary | 会话4185795% CI1.28% / 3.61%许可Proprietary | 详情 |
| 21 | Muse Wonder (Max)Meta | 2.43%净改善率 | 会话13368 | 95% CI1.08% / 3.78% | 许可Proprietary | 会话1336895% CI1.08% / 3.78%许可Proprietary | 来源 |
| 22 | DeepSeek-V4.1-Flash-20261017 (Max)DeepSeek | 2.37%净改善率 | 会话70887 | 95% CI1.80% / 2.94% | 许可MIT | 会话7088795% CI1.80% / 2.94%许可MIT | 来源 |
| 23 | GPT-6 Luna (Max)OpenAI | 1.89%净改善率 | 会话226528 | 95% CI1.45% / 2.33% | 许可Proprietary | 会话22652895% CI1.45% / 2.33%许可Proprietary | 来源 |
| 24 | GLM-5.3 (Max)Z.ai | 1.09%净改善率 | 会话96282 | 95% CI0.29% / 1.88% | 许可MIT | 会话9628295% CI0.29% / 1.88%许可MIT | 来源 |
| 25 | Kimi K3 (Max)Moonshot AI | 0.84%净改善率 | 会话161333 | 95% CI0.26% / 1.42% | 许可Kimi K3 license | 会话16133395% CI0.26% / 1.42%许可Kimi K3 license | 详情 |
| 26 | Hy4 previewTencent | 0.75%净改善率 | 会话101478 | 95% CI0.05% / 1.44% | 许可Apache 2.0 | 会话10147895% CI0.05% / 1.44%许可Apache 2.0 | 详情 |
| 27 | GLM-5.2 (Max)Z.ai | 0.64%净改善率 | 会话99315 | 95% CI-0.13% / 1.41% | 许可MIT | 会话9931595% CI-0.13% / 1.41%许可MIT | 来源 |
| 28 | Claude Haiku 5.5 (High)Anthropic | 0.62%净改善率 | 会话31204 | 95% CI-0.40% / 1.65% | 许可Proprietary | 会话3120495% CI-0.40% / 1.65%许可Proprietary | 来源 |
| 29 | Qwen3.8-MaxAlibaba | 0.48%净改善率 | 会话80210 | 95% CI-0.18% / 1.13% | 许可Proprietary | 会话8021095% CI-0.18% / 1.13%许可Proprietary | 来源 |
| 30 | Grok 4.6 (xHigh)xAI | 0.45%净改善率 | 会话30398 | 95% CI-0.72% / 1.63% | 许可Proprietary | 会话3039895% CI-0.72% / 1.63%许可Proprietary | 详情 |
| 31 | MiMo-V2.6-FlashXiaomi | 0.20%净改善率 | 会话30946 | 95% CI-0.93% / 1.33% | 许可MIT | 会话3094695% CI-0.93% / 1.33%许可MIT | 来源 |
| 32 | DeepSeek-V4-Pro (High, 20260813)DeepSeek | -0.30%净改善率 | 会话7708 | 95% CI-2.00% / 1.40% | 许可MIT | 会话770895% CI-2.00% / 1.40%许可MIT | 来源 |
| 33 | GPT-5.6 Terra (xHigh)OpenAI | -1.50%净改善率 | 会话44683 | 95% CI-2.29% / -0.72% | 许可Proprietary | 会话4468395% CI-2.29% / -0.72%许可Proprietary | 来源 |
| 34 | Gemini 3.7 Flash (High)Google | -1.60%净改善率 | 会话63739 | 95% CI-2.48% / -0.73% | 许可Proprietary | 会话6373995% CI-2.48% / -0.73%许可Proprietary | 详情 |
| 35 | GPT-6 Luna (Medium)OpenAI | -1.84%净改善率 | 会话67150 | 95% CI-2.37% / -1.31% | 许可Proprietary | 会话6715095% CI-2.37% / -1.31%许可Proprietary | 来源 |
| 36 | GLM-5.3-FlashZ.ai | -2.39%净改善率 | 会话95924 | 95% CI-3.09% / -1.70% | 许可MIT | 会话9592495% CI-3.09% / -1.70%许可MIT | 来源 |
| 37 | MiniMax-M3.1-FlashMiniMax | -2.84%净改善率 | 会话12516 | 95% CI-4.38% / -1.30% | 许可MiniMax Community License | 会话1251695% CI-4.38% / -1.30%许可MiniMax Community License | 来源 |
| 38 | Step 5 PreviewStepFun | -2.88%净改善率 | 会话16992 | 95% CI-4.27% / -1.48% | 许可Proprietary | 会话1699295% CI-4.27% / -1.48%许可Proprietary | 来源 |
| 39 | Qwen3.8-27BAlibaba | -4.28%净改善率 | 会话64285 | 95% CI-5.13% / -3.42% | 许可Apache 2.0 | 会话6428595% CI-5.13% / -3.42%许可Apache 2.0 | 来源 |
| 40 | Qwen3.8-Flash-NextAlibaba | -4.43%净改善率 | 会话110023 | 95% CI-5.16% / -3.70% | 许可qwen-community-1.0 | 会话11002395% CI-5.16% / -3.70%许可qwen-community-1.0 | 来源 |
| 41 | Gemini 3.1 Pro PreviewGoogle | -4.44%净改善率 | 会话88060 | 95% CI-5.86% / -3.03% | 许可Proprietary | 会话8806095% CI-5.86% / -3.03%许可Proprietary | 详情 |
| 42 | Beam 1.0 501B A23Breflection | -5.38%净改善率 | 会话12422 | 95% CI-7.89% / -2.88% | 许可Apache 2.0 | 会话1242295% CI-7.89% / -2.88%许可Apache 2.0 | 来源 |
| 43 | Mistral Large 4Mistral AI | -5.46%净改善率 | 会话5499 | 95% CI-7.39% / -3.53% | 许可Modified MIT | 会话549995% CI-7.39% / -3.53%许可Modified MIT | 来源 |
| 44 | GPT-6 Luna (Low)OpenAI | -5.93%净改善率 | 会话41812 | 95% CI-6.63% / -5.22% | 许可Proprietary | 会话4181295% CI-6.63% / -5.22%许可Proprietary | 来源 |
| 45 | Qwen3.7-MaxAlibaba | -6.52%净改善率 | 会话43307 | 95% CI-7.73% / -5.31% | 许可Proprietary | 会话4330795% CI-7.73% / -5.31%许可Proprietary | 来源 |
| 46 | Qwen3.7-PlusAlibaba | -7.46%净改善率 | 会话24577 | 95% CI-9.02% / -5.90% | 许可Proprietary | 会话2457795% CI-9.02% / -5.90%许可Proprietary | 来源 |
| 47 | Muse Spark 1.1Meta | -7.66%净改善率 | 会话150505 | 95% CI-8.32% / -7.00% | 许可Proprietary | 会话15050595% CI-8.32% / -7.00%许可Proprietary | 详情 |
| 48 | Muse Spark 1.2 (xHigh)Meta | -7.70%净改善率 | 会话55090 | 95% CI-8.54% / -6.85% | 许可Proprietary | 会话5509095% CI-8.54% / -6.85%许可Proprietary | 详情 |
| 49 | Hy3Tencent | -7.72%净改善率 | 会话32773 | 95% CI-9.17% / -6.27% | 许可Apache 2.0 | 会话3277395% CI-9.17% / -6.27%许可Apache 2.0 | 详情 |
| 50 | MiMo-V2.5-ProXiaomi | -9.03%净改善率 | 会话49644 | 95% CI-10.04% / -8.01% | 许可MIT | 会话4964495% CI-10.04% / -8.01%许可MIT | 来源 |