← Benchmarks
ScreenSpot
ScreenSpot is the first realistic GUI grounding benchmark that encompasses mobile, desktop, and web environments. The dataset comprises over 1,200 instructions from iOS, Android, macOS, Windows and Web environments, along with annotated element types (text and icon/widget), designed to evaluate visual GUI agents' ability to accurately locate screen elements based on natural language instructions.
id screenspot · max 1 · 16 models reported
| # | Model | Score |
|---|---|---|
| 1 | Qwen3 VL 32B Instruct Alibaba Cloud / Qwen Team · open | 0.96 |
| 2 | Qwen3 VL 32B Thinking Alibaba Cloud / Qwen Team · open | 0.96 |
| 3 | Qwen3 VL 235B A22B Instruct Alibaba Cloud / Qwen Team · open | 0.95 |
| 4 | Qwen3 VL 235B A22B Thinking Alibaba Cloud / Qwen Team · open | 0.95 |
| 5 | Qwen3 VL 30B A3B Instruct Alibaba Cloud / Qwen Team · open | 0.95 |
| 6 | Qwen3 VL 30B A3B Thinking Alibaba Cloud / Qwen Team · open | 0.95 |
| 7 | Qwen3 VL 8B Instruct Alibaba Cloud / Qwen Team · open | 0.94 |
| 8 | Qwen3 VL 4B Instruct Alibaba Cloud / Qwen Team · open | 0.94 |
| 9 | Qwen3 VL 8B Thinking Alibaba Cloud / Qwen Team · open | 0.94 |
| 10 | Qwen3 VL 4B Thinking Alibaba Cloud / Qwen Team · open | 0.93 |
| 11 | Qwen2.5 VL 32B Instruct Alibaba Cloud / Qwen Team · open | 0.89 |
| 12 | Nova 2 Pro Amazon | 0.88 |
| 13 | Qwen2.5 VL 72B Instruct Alibaba Cloud / Qwen Team · open | 0.87 |
| 14 | Nova 2 Omni Amazon | 0.85 |