Calibration

We-Math → Video-MME (w/o subtitle)

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

Video-MME (w/o subtitle) is estimated from We-Math with a Michaelis–Menten + offset curve fitted on 6 models measured on both: y = 0.0000 + 2.0000·x / (1.16859 + x), R² = 0.50, cross-validated error 8.9 pp. It is used for 0 estimates.

Fitted on

Model	We-Math	Video-MME (w/o subtitle)
Claude Opus 4.5	70.0%	81.4%
Gemini 3 Pro	86.9%	87.7%
GPT-5.2	79.0%	85.8%
Kimi K2.5	84.7%	83.2%
LFM3-VL-2B	68.6%	62.8%
Qwen3.5 397B	87.9%	83.7%