Calibration
Mystery Game Puzzles → FrontierPhysics
FrontierPhysics is estimated from Mystery Game Puzzles with a Hill curve fitted on 20 models measured on both: y = 0.0261 + (0.2851 − 0.0261)·x^6.00 / (0.67987^6.00 + x^6.00), R² = 0.78, cross-validated error 4.2 pp. It is used for 36 estimates.
| Estimated model | Mystery Game Puzzles | FrontierPhysics | Source |
|---|---|---|---|
| Qwen3.8 Max | 38.0% | 3.4% | estimated ± 4.2 pp, high confidence |
| Claude Sonnet 5 | 35.0% | 3.1% | estimated ± 4.2 pp, high confidence |
| DeepSeek V4 Flash 0731 | 34.0% | 3.0% | estimated ± 4.2 pp, high confidence |
| Gemini 3.5 Flash | 32.0% | 2.9% | estimated ± 4.2 pp, high confidence |
| Qwen3.7 Max | 32.0% | 2.9% | estimated ± 4.2 pp, high confidence |
| o3 | 29.0% | 2.8% | estimated ± 4.2 pp, high confidence |
| Muse Spark 1.3 | 25.0% | 2.7% | estimated ± 4.2 pp, high confidence |
| GPT-5 | 23.0% | 2.6% | estimated ± 4.2 pp, high confidence |
| Qwen3.6-35B-A3B | 22.0% | 2.6% | estimated ± 4.2 pp, high confidence |
| Gemini 3 Flash | 20.0% | 2.6% | estimated ± 4.2 pp, high confidence |
| Nemotron 3 Ultra | 20.0% | 2.6% | estimated ± 4.2 pp, high confidence |
| Qwen3.5 Flash | 20.0% | 2.6% | estimated ± 4.2 pp, high confidence |
| Gemini 3.5 Flash-Lite | 19.0% | 2.6% | estimated ± 4.2 pp, high confidence |
| Qwen 3.6 Max (preview) | 19.0% | 2.6% | estimated ± 4.2 pp, high confidence |
| Qwen3.5 397B | 18.0% | 2.6% | estimated ± 4.2 pp, high confidence |
| Claude Sonnet 4.5 | 17.0% | 2.6% | estimated ± 4.2 pp, high confidence |
| Qwen3.5-122B-A10B | 17.0% | 2.6% | estimated ± 4.2 pp, high confidence |
| Qwen3.5 Plus | 17.0% | 2.6% | estimated ± 4.2 pp, high confidence |
| Qwen3.7 Plus | 17.0% | 2.6% | estimated ± 4.2 pp, high confidence |
| Claude Sonnet 4.6 | 16.0% | 2.6% | estimated ± 4.2 pp, medium confidence |
| GPT-5.1 | 16.0% | 2.6% | estimated ± 4.2 pp, medium confidence |
| GLM-5.2 | 15.0% | 2.6% | estimated ± 4.2 pp, medium confidence |
| Qwen3.7 Flash | 15.0% | 2.6% | estimated ± 4.2 pp, medium confidence |
| GPT-4o mini | 12.0% | 2.6% | estimated ± 4.2 pp, medium confidence |
| Qwen3.6 Plus | 12.0% | 2.6% | estimated ± 4.2 pp, medium confidence |
| GLM-5.3-Flash | 8.0% | 2.6% | estimated ± 4.2 pp, medium confidence |
| MiniMax M3 | 8.0% | 2.6% | estimated ± 4.2 pp, medium confidence |
| GPT-4.1 mini | 7.0% | 2.6% | estimated ± 4.2 pp, medium confidence |
| GPT-5.4 mini | 7.0% | 2.6% | estimated ± 4.2 pp, medium confidence |
| o3-mini | 7.0% | 2.6% | estimated ± 4.2 pp, medium confidence |
| Qwen3.6-27B | 7.0% | 2.6% | estimated ± 4.2 pp, medium confidence |
| Inkling-Small | 6.0% | 2.6% | estimated ± 4.2 pp, medium confidence |
| GPT-5.4 nano | 5.0% | 2.6% | estimated ± 4.2 pp, medium confidence |
| o4-mini | 5.0% | 2.6% | estimated ± 4.2 pp, medium confidence |
| Qwen3 Max | 5.0% | 2.6% | estimated ± 4.2 pp, medium confidence |
| GPT-OSS 120B | 0.0% | 2.6% | estimated ± 4.2 pp, medium confidence |
Fitted on
| Model | Mystery Game Puzzles | FrontierPhysics |
|---|---|---|
| Claude Fable 5.1 | 58.0% | 11.3% |
| Claude Haiku 5.5 | 30.0% | 4.2% |
| Claude Opus 5 | 59.0% | 9.5% |
| Claude Opus 5.5 | 71.0% | 27.4% |
| DeepSeek V4.1 Flash 1017 | 49.0% | 3.6% |
| Gemini 3.8 Flash | 47.0% | 0.0% |
| Gemini 4 Argon | 64.0% | 8.3% |
| GLM-5.3 | 33.0% | 4.2% |
| GPT-5.6 Luna | 21.0% | 3.6% |
| GPT-5.6 Sol | 58.0% | 7.1% |
| GPT-5.6 Terra | 35.0% | 3.6% |
| GPT-6.1 Sol | 80.0% | 19.6% |
| GPT-6.1 Terra | 74.0% | 14.3% |
| GPT-6 Astra | 84.0% | 23.2% |
| GPT-6 Sol | 56.0% | 14.3% |
| Grok 4.6 | 34.0% | 2.4% |
| Grok 4.7 | 29.0% | 4.2% |
| Kimi K3.1 Pro | 42.0% | 6.5% |
| Kimi K3 | 26.0% | 2.4% |
| Mistral Large 4 | 17.0% | 0.0% |