ScreenSpot
multimodal
+
+
+
+
About
ScreenSpot is a GUI grounding benchmark that evaluates AI models' ability to locate and identify specific interface elements within screenshots of computer applications. This foundational evaluation tests visual understanding of graphical user interfaces, spatial reasoning for element localization, and the capacity to understand UI components across different software applications and operating systems.
+
+
+
+
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
88.5%
Average Score
86.8%
High Performers (80%+)
3Top Organizations
#1Alibaba Cloud / Qwen Team
3 models
86.8%
+
+
+
+
Leaderboard
3 models ranked by performance on ScreenSpot
| License | Links | ||||
|---|---|---|---|---|---|
| Feb 28, 2025 | Apache 2.0 | 88.5% | |||
| Jan 26, 2025 | tongyi-qianwen | 87.1% | |||
| Jan 26, 2025 | Apache 2.0 | 84.7% |
