Calibration
CursorBench 4.0 → VulcanBench v3
VulcanBench v3 is estimated from CursorBench 4.0 with a offset logistic curve fitted on 5 models measured on both: y = 0.6998 + (0.8700 − 0.6998) / (1 + exp(−175.03·(x − 0.3457))), R² = 1.00, cross-validated error 0.7 pp. It is used for 8 estimates.
| Estimated model | CursorBench 4.0 | VulcanBench v3 | Source |
|---|---|---|---|
| Claude Fable 5.1 | 51.8% | 87.0% | estimated ± 0.7 pp, low confidence |
| Claude Opus 5.5 | 57.8% | 87.0% | estimated ± 0.7 pp, low confidence |
| Claude Sonnet 5 | 34.1% | 75.2% | estimated ± 0.7 pp, low confidence |
| Claude Sonnet 5.5 | 55.5% | 87.0% | estimated ± 0.7 pp, low confidence |
| Composer 2.5 | 27.7% | 70.0% | estimated ± 0.7 pp, low confidence |
| Gemini 3.8 Flash | 39.6% | 87.0% | estimated ± 0.7 pp, medium confidence |
| Grok 4.7 | 46.3% | 87.0% | estimated ± 0.7 pp, medium confidence |
| Muse Spark 1.3 | 41.6% | 87.0% | estimated ± 0.7 pp, medium confidence |