benchgap
Calibration

CharXiv → ScreenSpot Pro

ScreenSpot Pro is estimated from CharXiv with a Hill curve fitted on 13 models measured on both: y = 0.0000 + (1.0714 − 0.0000)·x^6.00 / (0.71072^6.00 + x^6.00), R² = 0.76, cross-validated error 6.8 pp. It is used for 29 estimates.

Estimated modelCharXivScreenSpot ProSource
Claude Mythos 593.5%89.8%estimated ± 6.8 pp, medium confidence
Kimi K3.1 Pro93.4%89.7%estimated ± 6.8 pp, medium confidence
Glimpse92.9%89.3%estimated ± 6.8 pp, medium confidence
Kimi K391.3%87.6%estimated ± 6.8 pp, medium confidence
Claude Opus 4.791.0%87.3%estimated ± 6.8 pp, medium confidence
LongCat-2.590.2%86.5%estimated ± 6.8 pp, medium confidence
GLM-5.3-Flash89.4%85.5%estimated ± 6.8 pp, medium confidence
Gemini 3.7 Flash88.7%84.7%estimated ± 6.8 pp, medium confidence
Muse Spark 1.188.4%84.4%estimated ± 6.8 pp, medium confidence
Claude Sonnet 588.3%84.2%estimated ± 6.8 pp, medium confidence
Sakana Fugu-Ultra86.6%82.1%estimated ± 6.8 pp, medium confidence
Seed 2.1 Pro85.4%80.4%estimated ± 6.8 pp, medium confidence
Sakana Fugu85.1%80.0%estimated ± 6.8 pp, medium confidence
Gemini 3.5 Flash84.2%78.7%estimated ± 6.8 pp, medium confidence
MiniMax M3.1 Flash82.5%76.1%estimated ± 6.8 pp, medium confidence
Seed 2.1 Turbo82.5%76.1%estimated ± 6.8 pp, medium confidence
Inkling82.0%75.2%estimated ± 6.8 pp, medium confidence
Inkling-Small81.3%74.1%estimated ± 6.8 pp, medium confidence
MiMo-V2.581.0%73.6%estimated ± 6.8 pp, medium confidence
Nemotron 3.5 Super VL 120B A12B80.8%73.3%estimated ± 6.8 pp, medium confidence
Kimi K2.680.4%72.5%estimated ± 6.8 pp, medium confidence
Nemotron 3.5 Super VL 120B A12B NVFP479.9%71.6%estimated ± 6.8 pp, medium confidence
Qwen3.5-27B79.5%70.9%estimated ± 6.8 pp, medium confidence
Qwen3.5-35B-A3B77.5%67.2%estimated ± 6.8 pp, medium confidence
Claude Sonnet 4.677.4%67.0%estimated ± 6.8 pp, medium confidence
Qwen3.5-122B-A10B77.2%66.6%estimated ± 6.8 pp, medium confidence
Gemini 3.1 Flash-Lite73.2%58.3%estimated ± 6.8 pp, medium confidence
Grok 4.2060.9%30.4%estimated ± 6.8 pp, low confidence
Command A+52.7%15.3%estimated ± 6.8 pp, low confidence

Fitted on

ModelCharXivScreenSpot Pro
Claude Opus 4.568.5%45.7%
Claude Opus 4.889.9%87.9%
Gemini 3.1 Pro80.2%84.4%
Gemini 3 Pro81.4%72.7%
GPT-5.482.8%85.4%
Inkling 2 Small87.7%83.8%
Muse Glimmer 30B78.8%75.4%
Muse Spark86.4%84.1%
Nemotron 3 Nano Omni 30B A3B76.3%57.8%
Qwen3.5 397B80.8%65.6%
Qwen3.6 Plus81.5%68.2%
Qwen3.7 Plus85.9%79.0%
Qwen3.8 Max93.5%84.5%