benchgap
Calibration

OSWorld-Verified → JobBench

JobBench is estimated from OSWorld-Verified with a offset logistic curve fitted on 14 models measured on both: y = 0.3276 + (0.5215 − 0.3276) / (1 + exp(−60.16·(x − 0.7611))), R² = 0.60, cross-validated error 9.0 pp. It is used for 6 estimates.

Estimated modelOSWorld-VerifiedJobBenchSource
Holo3-122B-A10B78.9%49.0%estimated ± 9.0 pp, medium confidence
Holo3-35B-A3B82.6%51.8%estimated ± 9.0 pp, medium confidence
Qwen3.5-27B56.2%32.8%estimated ± 9.0 pp, medium confidence
Qwen3.5-35B-A3B54.5%32.8%estimated ± 9.0 pp, medium confidence
UI-Mate-27B77.0%45.0%estimated ± 9.0 pp, medium confidence
UI-Mate-9B66.2%32.8%estimated ± 9.0 pp, medium confidence