benchgap
Calibration

AA EnterpriseOps-Gym → AutomationBench

AutomationBench is estimated from AA EnterpriseOps-Gym with a linear curve fitted on 5 models measured on both: y = -0.9635·x + 0.8151, R² = 0.61, cross-validated error 7.0 pp. It is used for 8 estimates.

Estimated modelAA EnterpriseOps-GymAutomationBenchSource
Claude Fable 551.1%32.3%estimated ± 7.0 pp, low confidence
Gemini 3.5 Flash50.1%33.2%estimated ± 7.0 pp, low confidence
Grok 4.648.3%35.0%estimated ± 7.0 pp, low confidence
Inkling38.0%44.9%estimated ± 7.0 pp, low confidence
MiniMax M332.1%50.6%estimated ± 7.0 pp, low confidence
Muse Glimmer 30B34.7%48.1%estimated ± 7.0 pp, low confidence
Nemotron 3 Ultra28.9%53.7%estimated ± 7.0 pp, low confidence
Qwen3.8-27B44.2%38.9%estimated ± 7.0 pp, low confidence