模型多维对比
支持最多选择4个模型，进行各项指标的横向对比。
热门对决：
| 评估指标 | Claude Mythos Preview | Claude Fable 5 | Claude Fable 5.1 (XHigh) |
|---|---|---|---|
| 提供商 | Anthropic | Anthropic | Anthropic |
| 开源协议 | Proprietary | Proprietary | Proprietary |
| Podium 综合得分 | 97.4 | 93.4 | 89.0 |
| 综合排名 | 1 | 2 | — |
| 竞技场 Elo | — | 1509 | — |
| 智商指数 | — | 60.0 | — |
| 编程 | 80.9 | 84.1 | 88.0 |
| 数学 | 100.0 | — | — |
| 推理 | 98.8 | 81.0 | 85.0 |
| 智能体 | — | 96.2 | — |
| 知识 | 100.0 | 43.3 | 45.0 |
| 多模态 | 100.0 | — | 30.0 |
| 长上下文 | — | 99.2 | — |
| 速度 | — | 10.2 | 40.0 |
| 性价比 | — | 71.4 | 40.0 |
| 输出速度 | 80 t/s | 71 t/s | 78 t/s |
| 首Token延迟 | — | 141 ms | 350 ms |
| 输入价格 ($/M) | $3/M | $10/M | $3/M |
| 输出价格 ($/M) | $15/M | $50/M | $15/M |
| 上下文窗口 | 1000K | 1000K | 500K |
| SWE-Bench Verified | 93.9% | 95% | — |
| GPQA Diamond | 94.6% | 92.6% | — |
| Multilingual MMLU | 92.7% | — | — |
| Browsecomp | 86.9% | — | — |
| CharXiv Reasoning | 93.2% | — | — |
| Humanity’s Last Exam | 64.7% | 53.3% | — |
| SWE-Bench Pro | 77.8% | — | — |
| SciCode | — | 60.2% | — |
| LiveCodeBench | — | 70% | — |
| Ifbench | — | 63.5% | — |
| τ²-Bench Retail | — | 98.5% | — |
| Terminal-Bench Hard | — | 62.9% | — |