Calibration
Claw-Eval → VITA-Bench
VITA-Bench is estimated from Claw-Eval with a inverse Michaelis–Menten curve fitted on 7 models measured on both: y = 0.87521·(x − 0.0000) / (0.0000 + 2.0000 − x), R² = 0.38, cross-validated error 10.7 pp. It is used for 18 estimates.
| Estimated model | Claw-Eval | VITA-Bench | Source |
|---|---|---|---|
| Claude Sonnet 4.6 | 67.8% | 44.9% | estimated ± 10.7 pp, low confidence |
| Gemini 3.1 Pro | 57.8% | 35.6% | estimated ± 10.7 pp, low confidence |
| Gemini 3 Flash | 49.2% | 28.6% | estimated ± 10.7 pp, low confidence |
| GLM-5-Turbo | 55.8% | 33.9% | estimated ± 10.7 pp, low confidence |
| GLM-5V-Turbo | 53.8% | 32.2% | estimated ± 10.7 pp, low confidence |
| K-EXAONE 2.0 | 77.7% | 55.6% | estimated ± 10.7 pp, low confidence |
| LFM2.5-2.6B | 62.8% | 40.1% | estimated ± 10.7 pp, low confidence |
| LLaDA2.2-mini | 57.2% | 35.0% | estimated ± 10.7 pp, low confidence |
| MiMo-V2.5 | 62.3% | 39.6% | estimated ± 10.7 pp, low confidence |
| MiMo-V2.5-Pro | 63.8% | 41.0% | estimated ± 10.7 pp, low confidence |
| MiMo-V2-Omni | 45.2% | 25.6% | estimated ± 10.7 pp, low confidence |
| MiMo-V2-Pro | 57.8% | 35.6% | estimated ± 10.7 pp, low confidence |
| MiniMax M2.7 | 48.7% | 28.2% | estimated ± 10.7 pp, low confidence |
| Muse Spark | 63.8% | 41.0% | estimated ± 10.7 pp, low confidence |
| Nemotron 3 Super 100B | 5.5% | 2.5% | estimated ± 10.7 pp, low confidence |
| Ornith-1.0-35B | 69.8% | 46.9% | estimated ± 10.7 pp, low confidence |
| Ornith-1.0-397B | 77.1% | 54.9% | estimated ± 10.7 pp, low confidence |
| Ornith-1.0-9B | 63.1% | 40.3% | estimated ± 10.7 pp, low confidence |