Calibration
MathVision → MathVision w/ Python
MathVision w/ Python is estimated from MathVision with a Michaelis–Menten curve fitted on 5 models measured on both: y = 2.0000·x / (0.99256 + x), R² = 0.93, cross-validated error 0.5 pp. It is used for 14 estimates.
| Estimated model | MathVision | MathVision w/ Python | Source |
|---|---|---|---|
| Claude Opus 4.5 | 74.3% | 85.6% | estimated ± 0.5 pp, low confidence |
| dots3-note Preview | 87.7% | 93.8% | estimated ± 0.5 pp, low confidence |
| Gemini 3 Pro | 86.6% | 93.2% | estimated ± 0.5 pp, low confidence |
| Gemma 4 12B | 79.7% | 89.1% | estimated ± 0.5 pp, low confidence |
| GPT-5.2 | 83.0% | 91.1% | estimated ± 0.5 pp, low confidence |
| Kimi K2.6 | 87.4% | 93.6% | estimated ± 0.5 pp, low confidence |
| Qwen3.5-122B-A10B | 86.2% | 93.0% | estimated ± 0.5 pp, low confidence |
| Qwen3.5-27B | 86.0% | 92.8% | estimated ± 0.5 pp, low confidence |
| Qwen3.5-35B-A3B | 83.9% | 91.6% | estimated ± 0.5 pp, low confidence |
| Qwen3.5 397B | 88.6% | 94.3% | estimated ± 0.5 pp, low confidence |
| Qwen3.6 Plus | 88.0% | 94.0% | estimated ± 0.5 pp, low confidence |
| Qwen3.7 Plus | 90.3% | 95.3% | estimated ± 0.5 pp, medium confidence |
| Seed 2.1 Pro | 92.6% | 96.5% | estimated ± 0.5 pp, medium confidence |
| Seed 2.1 Turbo | 90.1% | 95.2% | estimated ± 0.5 pp, medium confidence |