Calibration
QwenClawBench → VITA-Bench
VITA-Bench is estimated from QwenClawBench with a Michaelis–Menten curve fitted on 6 models measured on both: y = 2.0000·x / (2.24842 + x), R² = 0.33, cross-validated error 8.2 pp. It is used for 4 estimates.
| Estimated model | QwenClawBench | VITA-Bench | Source |
|---|---|---|---|
| GLM-5 | 54.1% | 38.8% | estimated ± 8.2 pp, low confidence |
| Kimi K2.5 | 54.3% | 38.9% | estimated ± 8.2 pp, low confidence |
| Qwen3.6-27B | 53.4% | 38.4% | estimated ± 8.2 pp, low confidence |
| Qwen 3.6 Max (preview) | 59.0% | 41.6% | estimated ± 8.2 pp, low confidence |