| Rank | Model | Score | 95% rank interval | Submitted | AUD | SEM | VIS | AV | Human score | Rated cells | Human rank |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Claude Opus 5.5 | 92.22 | 1–3 | 156 | 95.0 | 96.6 | 86.3 | 89.4 | 88.64 | 77 | 1 |
| 2 | Gemini 4 Argon | 92.02 | 1–2 | 159 | 95.4 | 97.6 | 83.8 | 88.9 | 86.62 | 72 | 2 |
| 3 | GPT-6 Astra | 88.77 | 3–5 | 160 | 93.7 | 97.6 | 72.7 | 87.0 | 86.44 | 79 | 3 |
| 4 | DeepSeek V4.1 Flash | 88.34 | 3–5 | 156 | 91.8 | 95.1 | 76.5 | 86.5 | 80.07 | 62 | 6 |
| 5 | Doubao Seed 2.1 Pro | 86.87 | 4–6 | 160 | 92.8 | 94.2 | 71.6 | 85.0 | 82.85 | 79 | 4 |
| 6 | Qwen3.8 Max | 83.93 | 5–7 | 152 | 87.0 | 92.5 | 69.6 | 83.0 | 80.89 | 79 | 5 |
| 7 | Hy4 | 68.35 | 8–9 | 158 | 49.2 | 90.1 | 76.6 | 45.4 | 63.80 | 66 | 7 |
| 8 | MiMo v2.6 Pro | 63.35 | 8–9 | 120 | 67.3 | 71.0 | 51.8 | 60.9 | 59.32 | 78 | 8 |
Ranks are renumbered after omitting models excluded from this page; 95% intervals come from the original published board (paired bootstrap over tasks). AUD/SEM/VIS/AV are board means on the same 0–100 scale.