ScreenSpot-Pro: leaderboard
Professional GUI grounding benchmark requiring agents to identify precise screen locations in high-resolution development, creative, and scientific software.
Metric: Score (%). Source: llm-stats.com. Status: saturation imminent. 26 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-6 | 92.7 |
| 2 | Claude Opus 4.8 | 87.9 |
| 3 | GPT-5.2 | 86.3 |
| 4 | Qwen 3.8 Max | 84.5 |
| 5 | Muse Spark | 84.1 |
| 6 | Qwen 3.7 Plus | 79 |
| 7 | Muse Glimmer 30B | 75.4 |
| 8 | Gemini 3 Pro | 72.7 |
| 9 | Qwen 3.5 122B A10B | 70.4 |
| 10 | Qwen 3.5 27B | 70.3 |
| 11 | Gemini 3 Flash | 69.1 |
| 12 | Qwen 3.5 35B A3B | 68.6 |
| 13 | Qwen 3.6 Plus | 68.2 |
| 14 | Qwen 3 VL 235B A22B Instruct | 62 |
| 15 | Qwen 3 VL 235B A22B (Thinking) | 61.8 |
Interactive version: theaggregate.ai/benchmark?slug=screenspot-pro · How It Works · Data refreshed daily, snapshot 2026-09-05.