| # | Model | Net improvement | Samples |
|---|---|---|---|
| 1 |  Claude Fable 5.1 (Max)Anthropic | 13.85%±1.92% | 12,416 |
| 2 |  GPT 6 Astra (Max)OpenAI | 12.39%±2.6% | 9,692 |
| 3 | Claude Opus 5 (High)Anthropic | 11.06%±1.7% | 23,679 |
| 4 | Claude Opus 5 (Max)Anthropic | 10.8%±1.88% | 19,080 |
| 5 | Claude Fable 5 (High)Anthropic | 9.09%±1.42% | 36,950 |
| 6 | Claude Opus 4.8 (High)Anthropic | 7.57%±1.4% | 39,316 |
| 7 |
| GPT 5.6 Sol (xHigh)OpenAI |
|---|
| 7.51%±1.42% |
| 30,881 |
| 8 |  Kimi K3 (Max)Moonshot | 6.46%±0.7% | 101,022 |
|---|
| 9 | Claude Sonnet 5 (High)Anthropic | 5.88%±1.82% | 29,782 |
|---|
| 10 | Hy4 previewTencent | 5.23%±1.31% | 19,613 |
|---|
| 11 | GPT 5.5 (xHigh)OpenAI | 5.22%±1.08% | 52,336 |
|---|
| 12 |  GLM 5.2 (Max)Z.ai | 4.56%±0.74% | 73,382 |
|---|
| 13 | 4.23%±1.77% | 8,770 |
| 14 |  DeepSeek V4 Pro (High) (0813)DeepSeek | 4.14%±0.92% | 39,047 |
|---|
| 15 | Qwen3.8 MaxAlibaba | 3.97%±1.02% | 23,261 |
|---|
| 16 |  Gemini 3.8 Flash (High)Google | 3.84%±1.96% | 10,105 |
|---|
| 17 | Grok 4.5SpaceXAI | 3.83%±1.13% | 36,452 |
|---|
| 18 | Grok 4.6 (xHigh)SpaceXAI | 3.12%±1.14% | 20,911 |
|---|
| 19 | GPT 5.5OpenAI | 2.8%±0.93% | 80,458 |
|---|
| 20 | GLM 5.3 (Max)Z.ai | 2.62%±0.66% | 65,488 |
|---|
Dots: score · lines: published uncertainty