VG-GUI-Bench (10 Uniform Frames): leaderboard
Metric: Step accuracy (%; 0.3 for the right action type plus up to 0.7 for its arguments, over the 1,071 steps of 100 mobile GUI episodes; ten uniformly sampled tutorial frames). Source: arxiv.org. Saturation forecast: Around January 2027. 7 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 61.68 |
| 2 | GPT-5 Mini | 58.4 |
| 3 | Kimi K2.5 | 58.22 |
| 4 | Claude Sonnet 4.6 | 45.8 |
| 5 | Seed 2.0 Pro | 39.78 |
| 6 | Qwen 3 VL 235B A22B | 26.88 |
Interactive version: theaggregate.ai/benchmark?slug=vg-gui-bench-10-uniform-frames · How It Works · Data refreshed daily, snapshot 2026-09-26.