Claude Opus 5.5 (Max)
Anthropic
18.63%
净改善率
评估 AI 在真实任务中的执行表现。分数表示任务效果的相对改善幅度,数值越高,代表改善越明显。
Anthropic
18.63%
净改善率
Anthropic
27.42%
净改善率
谷歌
18.41%
净改善率
| 排名 | 模型 | 机构 | 净改善率 | 置信区间 | 会话数 |
|---|---|---|---|---|---|
| 1 | Claude Fable 5.2 (Max) | Anthropic | 27.42% | 24.46% – 30.38% | 12,330 |
| 2 | Claude Opus 5.5 (Max) | Anthropic | 18.63% | 16.60% – 20.65% | 14,920 |
| 3 | Gemini 4 Neon (High) | 谷歌 | 18.41% | 15.90% – 20.93% | 5,696 |
| 4 | Gemini 4 Carbon (High) | 谷歌 | 16.10% | 15.41% – 16.79% | 64,114 |
| 5 | Claude Sonnet 5.5 (Max) | Anthropic | 15.56% | 12.66% – 18.46% | 8,912 |
| 6 | GPT-6.2 Astra (Max) | OpenAI | 15.22% | 12.48% – 17.96% | 4,785 |
| 7 | GPT-6 Astra (Max) | OpenAI | 14.56% | 12.53% – 16.59% | 15,818 |
| 8 | Claude Opus 5.5 (High) | Anthropic | 13.70% | 12.64% – 14.76% | 24,372 |
| 9 | Claude Fable 5.1 (Max) | Anthropic | 13.66% | 11.58% – 15.74% | 17,527 |
| 10 | Gemini 4 Hydrogen (High) | 谷歌 | 13.52% | 11.76% – 15.28% | 9,133 |
| 11 | GPT-6.2 Sol (Max) | OpenAI | 11.31% | 8.89% – 13.73% | 5,360 |
| 12 | GPT-6.1 Sol (Max) | OpenAI | 11.31% | 10.20% – 12.42% | 31,474 |
| 13 | GPT-6.1 Terra (Max) | OpenAI | 10.48% | 9.31% – 11.66% | 28,135 |
| 14 | Gemini 4 Argon (High) | 谷歌 | 10.31% | 9.39% – 11.23% | 55,384 |
| 15 | GPT-6.1 Sol (Medium) | OpenAI | 8.20% | 7.20% – 9.21% | 23,572 |
| 16 | Muse Wonder Full (Max) | Meta | 8.00% | 7.05% – 8.95% | 31,934 |
| 17 | Claude Sonnet 5.5 (High) | Anthropic | 7.42% | 5.47% – 9.38% | 11,317 |
| 18 | Qwen4-Max (xHigh) | 阿里巴巴 | 7.27% | 5.92% – 8.61% | 14,578 |
| 19 | GPT-6 Sol (Max) | OpenAI | 6.47% | 4.93% – 8.00% | 13,412 |
| 20 | Claude Opus 5 (Max) | Anthropic | 6.26% | 4.96% – 7.57% | 32,019 |
| 21 | Hy4 | 腾讯 | 6.09% | 4.90% – 7.29% | 27,366 |
| 22 | Claude Fable 5 (High) | Anthropic | 5.86% | 4.65% – 7.07% | 51,153 |
| 23 | GPT-6.1 Terra (Medium) | OpenAI | 4.61% | 4.04% – 5.18% | 54,799 |
| 24 | Kimi K3.1 Pro (Max) | 月之暗面 | 3.88% | 2.76% – 5.00% | 22,130 |
| 25 | GPT-5.6 Sol (xHigh) | OpenAI | 3.68% | 2.58% – 4.79% | 45,448 |
| 26 | Inkling 2 Small | thinky | 1.51% | 0.51% – 2.51% | 22,258 |
| 27 | Gemini 3.8 Flash (High) | 谷歌 | 1.49% | 0.57% – 2.41% | 52,549 |
| 28 | Grok 4.7 (xHigh) | xai | 1.48% | -0.04% – 3.00% | 19,072 |
| 29 | MiMo-V2.6-Pro | 小米 | 1.42% | 0.52% – 2.32% | 53,011 |
| 30 | Muse Spark 1.3 (Max) | Meta | 1.11% | 0.23% – 2.00% | 80,937 |
| 31 | Claude Sonnet 5 (High) | Anthropic | 1.00% | -0.20% – 2.21% | 41,853 |
| 32 | Muse Ember (Max) | Meta | 0.96% | 0.04% – 1.89% | 40,525 |
| 33 | DeepSeek-V4.1-Flash-20261017 (Max) | 深度求索 | 0.78% | 0.29% – 1.28% | 98,592 |
| 34 | GPT-6 Luna (Max) | OpenAI | 0.64% | 0.17% – 1.12% | 237,330 |
| 35 | Muse Wonder (Max) | Meta | 0.58% | -0.71% – 1.88% | 15,313 |
| 36 | GLM-5.3 (Max) | zai | -0.04% | -0.91% – 0.83% | 95,055 |
| 37 | Kimi K3 (Max) | 月之暗面 | -0.15% | -0.79% – 0.50% | 160,339 |
| 38 | GLM-5.2 (Max) | zai | -0.41% | -1.24% – 0.42% | 99,249 |
| 39 | Hy4 preview | 腾讯 | -0.44% | -1.18% – 0.30% | 104,712 |
| 40 | Grok 4.6 (xHigh) | xai | -0.60% | -1.89% – 0.70% | 30,266 |
| 41 | Claude Haiku 5.5 (High) | Anthropic | -0.61% | -1.65% – 0.42% | 33,048 |
| 42 | Qwen3.8-Max | 阿里巴巴 | -0.91% | -1.62% – -0.21% | 81,929 |
| 43 | MiMo-V2.6-Flash | 小米 | -1.75% | -2.85% – -0.66% | 37,071 |
| 44 | GPT-5.6 Terra (xHigh) | OpenAI | -2.67% | -3.51% – -1.83% | 44,415 |
| 45 | GPT-6 Luna (Medium) | OpenAI | -3.05% | -3.49% – -2.61% | 98,306 |
| 46 | Gemini 3.7 Flash (High) | 谷歌 | -3.07% | -4.01% – -2.13% | 63,059 |
| 47 | MiniMax-M3.1-Flash | MiniMax | -3.67% | -5.11% – -2.22% | 14,765 |
| 48 | GLM-5.3-Flash | zai | -3.87% | -4.61% – -3.12% | 95,827 |
| 49 | Qwen3.8-Flash-Next | 阿里巴巴 | -4.91% | -5.69% – -4.13% | 111,029 |
| 50 | Step 5 Preview | 阶跃星辰 | -5.15% | -6.56% – -3.73% | 17,306 |
| 51 | Qwen3.8-27B | 阿里巴巴 | -5.50% | -6.44% – -4.56% | 63,880 |
| 52 | Gemini 3.1 Pro Preview | 谷歌 | -5.69% | -7.44% – -3.95% | 87,229 |
| 53 | GPT-6 Luna (Low) | OpenAI | -6.39% | -7.08% – -5.70% | 49,072 |
| 54 | Mistral Large 4 | mistral | -7.15% | -9.19% – -5.11% | 5,843 |
| 55 | Qwen3.7-Max | 阿里巴巴 | -8.25% | -9.56% – -6.94% | 43,149 |
| 56 | Beam 1.0 501B A23B | reflection | -8.34% | -11.24% – -5.44% | 12,256 |
| 57 | Hy3 | 腾讯 | -8.63% | -10.17% – -7.10% | 32,871 |
| 58 | Qwen3.7-Plus | 阿里巴巴 | -8.98% | -10.59% – -7.38% | 24,745 |
| 59 | Muse Spark 1.1 | Meta | -9.05% | -9.72% – -8.37% | 147,881 |
| 60 | Muse Spark 1.2 (xHigh) | Meta | -9.11% | -10.03% – -8.19% | 54,656 |
| 61 | MiniMax-M3 | MiniMax | -10.92% | -12.10% – -9.74% | 44,442 |
| 62 | Spark-X2.6 | synlan | -12.35% | -13.56% – -11.14% | 15,913 |
| 63 | Mistral Medium 3.5 | mistral | -14.46% | -16.25% – -12.66% | 13,863 |
| 64 | Nemotron 3 Ultra | nvidia | -16.80% | -20.17% – -13.43% | 13,818 |
| 65 | NVIDIA-Nemotron-3.5-Super-VL-120B-A12B (NVFP4) | nvidia | -17.34% | -18.24% – -16.44% | 40,807 |
| 66 | Inkling | thinky | -18.45% | -20.05% – -16.86% | 43,733 |
| 67 | Inkling-Small | thinky | -18.53% | -20.59% – -16.47% | 14,689 |
| 68 | Solar Pro 4 | upstage | -23.89% | -26.15% – -21.63% | 11,137 |
数据来源:Arena 官方总和评估。 使用官方公开数据集,按 CC BY 4.0 标注来源。 各榜单独同步;来源未发布新数据时,内容日期保持不变。
净改善率反映智能体任务中的相对改进,不是综合能力百分制。 最近检查:2026-11-22 18:01(UTC+8)。