benchgap
Calibration

MMMU-Pro w/ Python → AA-MMMU-Pro

AA-MMMU-Pro is estimated from MMMU-Pro w/ Python with a Hill curve fitted on 10 models measured on both: y = 0.1825 + (0.9318 − 0.1825)·x^6.00 / (0.63760^6.00 + x^6.00), R² = 0.93, cross-validated error 2.0 pp. It is used for 0 estimates.

Fitted on

ModelMMMU-Pro w/ PythonAA-MMMU-Pro
GPT-5.482.1%78.4%
GPT-5.4 mini78.0%73.3%
GPT-5.4 nano69.5%65.4%
GPT-5.583.2%79.9%
GPT-5.6 Luna79.5%78.6%
GPT-5.6 Sol84.6%83.4%
GPT-5.6 Terra82.0%80.7%
Kimi K2.680.1%79.4%
Kimi K3.1 Pro90.0%83.9%
Kimi K383.4%80.5%