Calibration

ARC-AGI-1 → HLE w/o tools

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

HLE w/o tools is estimated from ARC-AGI-1 with a offset logistic curve fitted on 12 models measured on both: y = 0.2820 + (1.2000 − 0.2820) / (1 + exp(−27.79·(x − 0.9995))), R² = 0.86, cross-validated error 6.0 pp. It is used for 1 estimate.

Estimated model	ARC-AGI-1	HLE w/o tools	Source
Claude Sonnet 4.5 Thinking	63.7%	28.2%	estimated ± 6.0 pp, medium confidence