Calibration

VideoMMMU → ERQA

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

ERQA is estimated from VideoMMMU with a Michaelis–Menten + offset curve fitted on 8 models measured on both: y = 0.1321 + 2.0000·x / (2.44830 + x), R² = 0.36, cross-validated error 9.6 pp. It is used for 0 estimates.

Fitted on

Model	VideoMMMU	ERQA
Claude Opus 4.5	84.4%	46.8%
Gemini 3 Pro	87.6%	70.5%
GPT-5.2	85.9%	59.8%
LFM3-VL-2B	49.6%	48.2%
Qwen3.5 397B	84.7%	67.5%
Qwen3.6-27B	84.4%	62.5%
Qwen3.7 Plus	85.4%	69.8%
Qwen3.8 Max	88.7%	77.8%