benchgap
Calibration

EnterpriseOps-Gym → Harvey LAB

Harvey LAB is estimated from EnterpriseOps-Gym with a Hill curve fitted on 6 models measured on both: y = 0.0000 + (0.9460 − 0.0000)·x^6.00 / (0.21081^6.00 + x^6.00), R² = 0.98, cross-validated error 1.2 pp. It is used for 9 estimates.

Estimated modelEnterpriseOps-GymHarvey LABSource
GLM-5.3-Flash33.2%88.8%estimated ± 1.2 pp, medium confidence
GLM-5.3 (max)36.4%91.2%estimated ± 1.2 pp, medium confidence
Qwen3.8 27B (xhigh)44.2%93.5%estimated ± 1.2 pp, medium confidence
Inkling (xhigh)38.0%91.9%estimated ± 1.2 pp, medium confidence
Muse Glimmer (high)34.7%90.1%estimated ± 1.2 pp, medium confidence
Grok 4.6 (high)48.3%93.9%estimated ± 1.2 pp, medium confidence
Gemini 3.7 Flash (medium)50.4%94.1%estimated ± 1.2 pp, medium confidence
Gemini 3.5 Flash (high)50.1%94.1%estimated ± 1.2 pp, medium confidence
DeepSeek V4 Pro 0813 (max)49.6%94.0%estimated ± 1.2 pp, medium confidence