Calibration

AA Terminal-Bench 2.1 → Terminal-Bench 2.1 (Vals AI)

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

Terminal-Bench 2.1 (Vals AI) is estimated from AA Terminal-Bench 2.1 with a offset logistic curve fitted on 11 models measured on both: y = 0.5106 + (0.8800 − 0.5106) / (1 + exp(−45.90·(x − 0.8375))), R² = 0.92, cross-validated error 6.0 pp. It is used for 1 estimate.

Estimated model	AA Terminal-Bench 2.1	Terminal-Bench 2.1 (Vals AI)	Source
Muse Glimmer 30B	51.7%	51.1%	estimated ± 6.0 pp, low confidence