Calibration

ARC-AGI-2 → HLE w/o tools

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

HLE w/o tools is estimated from ARC-AGI-2 with a linear curve fitted on 16 models measured on both: y = 0.3480·x + 0.2058, R² = 0.68, cross-validated error 6.9 pp. It is used for 1 estimate.

Estimated model	ARC-AGI-2	HLE w/o tools	Source
Gemini 3 Pro Deep Think	45.1%	36.3%	estimated ± 6.9 pp, medium confidence