Calibration
MathVision → PerceptionBench
PerceptionBench is estimated from MathVision with a inverse Michaelis–Menten curve fitted on 5 models measured on both: y = 0.70017·(x − 0.0000) / (0.0000 + 2.0000 − x), R² = 0.58, cross-validated error 3.6 pp. It is used for 12 estimates.
| Estimated model | MathVision | PerceptionBench | Source |
|---|---|---|---|
| Qwen3.8-Omni-Flash | 91.8% | 59.4% | estimated ± 3.6 pp, medium confidence |
| Qwen3.8-Flash-Next | 90.6% | 58.0% | estimated ± 3.6 pp, medium confidence |
| Qwen3.7 Plus | 90.3% | 57.6% | estimated ± 3.6 pp, medium confidence |
| Qwen3.5 397B | 88.6% | 55.7% | estimated ± 3.6 pp, medium confidence |
| Qwen3.6 Plus | 88.0% | 55.0% | estimated ± 3.6 pp, medium confidence |
| Kimi K2.6 | 87.4% | 54.3% | estimated ± 3.6 pp, low confidence |
| Gemini 3 Pro | 86.6% | 53.5% | estimated ± 3.6 pp, low confidence |
| Qwen3.5-122B-A10B | 86.2% | 53.0% | estimated ± 3.6 pp, low confidence |
| Qwen3.5-27B | 86.0% | 52.8% | estimated ± 3.6 pp, low confidence |
| Kimi K2.5 | 84.2% | 50.9% | estimated ± 3.6 pp, low confidence |
| Qwen3.5-35B-A3B | 83.9% | 50.6% | estimated ± 3.6 pp, low confidence |
| GPT-5.2 | 83.0% | 49.7% | estimated ± 3.6 pp, low confidence |
Fitted on
| Model | MathVision | PerceptionBench |
|---|---|---|
| dots3-note Preview | 87.7% | 53.4% |
| Inkling 2 Mini | 89.0% | 57.5% |
| Inkling 2 Small | 92.8% | 64.7% |
| Kimi K3 | 94.3% | 58.5% |
| Qwen3.8 Max | 95.2% | 63.5% |