benchgap
Calibration

Video-MME (with subtitle) → ScreenSpot Pro

ScreenSpot Pro is estimated from Video-MME (with subtitle) with a Hill curve fitted on 5 models measured on both: y = 0.0000 + (1.2000 − 0.0000)·x^6.00 / (0.81765^6.00 + x^6.00), R² = 0.83, cross-validated error 6.9 pp. It is used for 2 estimates.

Estimated modelVideo-MME (with subtitle)ScreenSpot ProSource
MiMo-V2.587.7%72.4%estimated ± 6.9 pp, low confidence
MiniMax M385.4%67.8%estimated ± 6.9 pp, low confidence