ScreenSpot Pro
multimodal
+
+
+
+
About
ScreenSpot-Pro is an advanced GUI grounding benchmark designed for professional high-resolution environments, covering 23 applications across five industries and three operating systems. This rigorous evaluation reveals significant challenges in current multimodal models, with the best achieving only 18.9% accuracy, highlighting the complexity of professional GUI understanding in authentic high-resolution settings.
+
+
+
+
Evaluation Stats
Total Models3
Organizations1
Verified Results0
Self-Reported3
+
+
+
+
Benchmark Details
Max Score1
Language
en
+
+
+
+
Performance Overview
Score distribution and top performers
Score Distribution
3 models
Top Score
43.6%
Average Score
37.3%
High Performers (80%+)
0Top Organizations
#1Alibaba Cloud / Qwen Team
3 models
37.3%
+
+
+
+
Leaderboard
3 models ranked by performance on ScreenSpot Pro
| License | Links | ||||
|---|---|---|---|---|---|
| Jan 26, 2025 | tongyi-qianwen | 43.6% | |||
| Feb 28, 2025 | Apache 2.0 | 39.4% | |||
| Jan 26, 2025 | Apache 2.0 | 29.0% |
