Calibration

VideoMMMU → We-Math

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

We-Math is estimated from VideoMMMU with a Michaelis–Menten + offset curve fitted on 6 models measured on both: y = 0.4539 + 2.0000·x / (3.85891 + x), R² = 0.43, cross-validated error 8.7 pp. It is used for 0 estimates.

Fitted on

Model	VideoMMMU	We-Math
Claude Opus 4.5	84.4%	70.0%
Gemini 3 Pro	87.6%	86.9%
GPT-5.2	85.9%	79.0%
Kimi K2.5	86.6%	84.7%
LFM3-VL-2B	49.6%	68.6%
Qwen3.5 397B	84.7%	87.9%