benchgap
Calibration

ZeroBench → MMMU-Pro

MMMU-Pro is estimated from ZeroBench with a offset logistic curve fitted on 9 models measured on both: y = 0.8027 + (0.8241 − 0.8027) / (1 + exp(−200.00·(x − 0.2274))), R² = 0.38, cross-validated error 1.7 pp. It is used for 0 estimates.

Fitted on

ModelZeroBenchMMMU-Pro
dots3-note Preview19.0%79.1%
Gemini 3.1 Pro29.0%83.9%
GPT-5.441.0%81.2%
Kimi K3.1 Pro39.0%84.1%
Kimi K323.0%81.6%
Muse Spark33.0%80.4%
Qwen3.8 Max24.0%82.3%
Seed 2.1 Pro18.0%81.6%
Seed 2.1 Turbo11.0%80.1%