Calibration
Mystery Game Puzzles → MaxxBench
MaxxBench is estimated from Mystery Game Puzzles with a Hill curve fitted on 13 models measured on both: y = 0.4718 + (0.7734 − 0.4718)·x^6.00 / (0.54996^6.00 + x^6.00), R² = 0.65, cross-validated error 8.8 pp. It is used for 20 estimates.
| Estimated model | Mystery Game Puzzles | MaxxBench | Source |
|---|---|---|---|
| o3 | 29.0% | 47.8% | estimated ± 8.8 pp, medium confidence |
| Qwen3.5 Flash | 20.0% | 47.3% | estimated ± 8.8 pp, medium confidence |
| Qwen 3.6 Max (preview) | 19.0% | 47.2% | estimated ± 8.8 pp, medium confidence |
| Qwen3.5 397B | 18.0% | 47.2% | estimated ± 8.8 pp, medium confidence |
| Claude Sonnet 4.5 | 17.0% | 47.2% | estimated ± 8.8 pp, medium confidence |
| Qwen3.5-122B-A10B | 17.0% | 47.2% | estimated ± 8.8 pp, medium confidence |
| Qwen3.5 Plus | 17.0% | 47.2% | estimated ± 8.8 pp, medium confidence |
| Qwen3.7 Plus | 17.0% | 47.2% | estimated ± 8.8 pp, medium confidence |
| Claude Sonnet 4.6 | 16.0% | 47.2% | estimated ± 8.8 pp, medium confidence |
| GPT-5.1 | 16.0% | 47.2% | estimated ± 8.8 pp, medium confidence |
| Qwen3.7 Flash | 15.0% | 47.2% | estimated ± 8.8 pp, medium confidence |
| GPT-4o mini | 12.0% | 47.2% | estimated ± 8.8 pp, medium confidence |
| Qwen3.6 Plus | 12.0% | 47.2% | estimated ± 8.8 pp, medium confidence |
| GPT-4.1 mini | 7.0% | 47.2% | estimated ± 8.8 pp, medium confidence |
| GPT-5.4 mini | 7.0% | 47.2% | estimated ± 8.8 pp, medium confidence |
| o3-mini | 7.0% | 47.2% | estimated ± 8.8 pp, medium confidence |
| GPT-5.4 nano | 5.0% | 47.2% | estimated ± 8.8 pp, low confidence |
| o4-mini | 5.0% | 47.2% | estimated ± 8.8 pp, low confidence |
| Qwen3 Max | 5.0% | 47.2% | estimated ± 8.8 pp, low confidence |
| GPT-OSS 120B | 0.0% | 47.2% | estimated ± 8.8 pp, low confidence |
Fitted on
| Model | Mystery Game Puzzles | MaxxBench |
|---|---|---|
| Claude Fable 5 | 52.0% | 63.5% |
| Claude Opus 5 | 59.0% | 63.5% |
| Claude Opus 5.5 | 71.0% | 76.4% |
| Claude Sonnet 5.5 | 65.0% | 66.9% |
| DeepSeek V4.1 Flash 1017 | 49.0% | 55.6% |
| DeepSeek V4 Flash 0731 | 34.0% | 28.1% |
| Gemini 3.8 Flash | 47.0% | 53.4% |
| GLM-5.3 | 33.0% | 56.7% |
| GLM-5.3-Flash | 8.0% | 55.1% |
| GPT-5.6 Luna | 21.0% | 46.1% |
| GPT-5.6 Sol | 58.0% | 69.1% |
| GPT-6 Astra | 84.0% | 71.9% |
| GPT-6 Luna | 7.0% | 51.7% |