Calibration

V* → MLVU (M-Avg)

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap
License: https://creativecommons.org/licenses/by-nc/4.0/

MLVU (M-Avg) is estimated from V* with a offset logistic curve fitted on 6 models measured on both: y = 0.7972 + (0.8670 − 0.7972) / (1 + exp(−200.00·(x − 0.8830))), R² = 0.34, cross-validated error 6.4 pp. It is used for 0 estimates.

Fitted on

Model	V*	MLVU (M-Avg)
Claude Opus 4.5	67.0%	81.7%
Gemini 3 Pro	88.0%	83.0%
GPT-5.2	75.9%	85.6%
LFM3-VL-2B	86.4%	71.3%
Qwen3.5 397B	95.8%	86.7%
Qwen3.6-27B	94.7%	86.6%