Calibration

PostTrainBench v1.1 → AA Coding Index

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

AA Coding Index is estimated from PostTrainBench v1.1 with a linear curve fitted on 11 models measured on both: y = 0.3803·x + 0.6269, R² = 0.54, cross-validated error 3.1 pp. It is used for 1 estimate.

Estimated model	PostTrainBench v1.1	AA Coding Index	Source
Claude Opus 4.7	28.6%	73.5%	estimated ± 3.1 pp, high confidence