benchgap
Calibration

MathVision → ScreenSpot Pro

ScreenSpot Pro is estimated from MathVision with a linear curve fitted on 6 models measured on both: y = 1.8439·x + -0.9144, R² = 0.91, cross-validated error 4.5 pp. It is used for 12 estimates.

Estimated modelMathVisionScreenSpot ProSource
Gemma 4 12B79.7%55.5%estimated ± 4.5 pp, medium confidence
GPT-5.283.0%61.6%estimated ± 4.5 pp, medium confidence
Kimi K2.687.4%69.7%estimated ± 4.5 pp, medium confidence
Kimi K394.3%82.4%estimated ± 4.5 pp, medium confidence
Qwen3.5-122B-A10B86.2%67.5%estimated ± 4.5 pp, medium confidence
Qwen3.5-27B86.0%67.1%estimated ± 4.5 pp, medium confidence
Qwen3.5-35B-A3B83.9%63.3%estimated ± 4.5 pp, medium confidence
Qwen3.8-27B90.0%74.5%estimated ± 4.5 pp, medium confidence
Qwen3.8-Flash-Next90.6%75.6%estimated ± 4.5 pp, medium confidence
Qwen3.8-Omni-Flash91.8%77.8%estimated ± 4.5 pp, medium confidence
Seed 2.1 Pro92.6%79.3%estimated ± 4.5 pp, medium confidence
Seed 2.1 Turbo90.1%74.7%estimated ± 4.5 pp, medium confidence