Calibration
CharXiv → MMVU
MMVU is estimated from CharXiv with a Michaelis–Menten + offset curve fitted on 6 models measured on both: y = 0.2844 + 2.0000·x / (2.54828 + x), R² = 0.54, cross-validated error 3.6 pp. It is used for 14 estimates.
| Estimated model | CharXiv | MMVU | Source |
|---|---|---|---|
| Claude Mythos 5 | 93.5% | 82.1% | estimated ± 3.6 pp, medium confidence |
| Glimpse | 92.9% | 81.9% | estimated ± 3.6 pp, medium confidence |
| Claude Opus 4.7 | 91.0% | 81.1% | estimated ± 3.6 pp, medium confidence |
| LongCat-2.5 | 90.2% | 80.7% | estimated ± 3.6 pp, medium confidence |
| Claude Opus 4.8 | 89.9% | 80.6% | estimated ± 3.6 pp, medium confidence |
| Gemini 3.7 Flash | 88.7% | 80.1% | estimated ± 3.6 pp, medium confidence |
| Muse Spark 1.1 | 88.4% | 79.9% | estimated ± 3.6 pp, medium confidence |
| Claude Sonnet 5 | 88.3% | 79.9% | estimated ± 3.6 pp, medium confidence |
| Sakana Fugu-Ultra | 86.6% | 79.2% | estimated ± 3.6 pp, medium confidence |
| Sakana Fugu | 85.1% | 78.5% | estimated ± 3.6 pp, medium confidence |
| MiniMax M3.1 Flash | 82.5% | 77.4% | estimated ± 3.6 pp, medium confidence |
| Claude Sonnet 4.6 | 77.4% | 75.0% | estimated ± 3.6 pp, medium confidence |
| Nemotron 3 Nano Omni 30B A3B | 76.3% | 74.5% | estimated ± 3.6 pp, low confidence |
| Gemini 3.1 Flash-Lite | 73.2% | 73.1% | estimated ± 3.6 pp, low confidence |
Fitted on
| Model | CharXiv | MMVU |
|---|---|---|
| GLM-5.3-Flash | 89.4% | 80.5% |
| Kimi K2.5 | 77.5% | 80.4% |
| Qwen3.5-122B-A10B | 77.2% | 74.7% |
| Qwen3.5-27B | 79.5% | 73.3% |
| Qwen3.5-35B-A3B | 77.5% | 72.3% |
| Qwen3.8 Max | 93.5% | 82.4% |