benchgap
Calibration

ARC-AGI-1 → Artificial Analysis Intelligence Index

Artificial Analysis Intelligence Index is estimated from ARC-AGI-1 with a Hill curve fitted on 32 models measured on both: y = 0.1834 + (1.2000 − 0.1834)·x^6.00 / (1.11705^6.00 + x^6.00), R² = 0.63, cross-validated error 7.0 pp. It is used for 1 estimate.

Estimated modelARC-AGI-1Artificial Analysis Intelligence IndexSource
Claude Sonnet 4.5 Thinking63.7%21.7%estimated ± 7.0 pp, medium confidence