AutoGUI-v2 - Element Captioning: leaderboard
Metric: Element interaction-outcome captioning on 788 tasks, multiple-choice accuracy (%) with easy and hard (same-group) distractors; the authors' tuned prompts; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro | 70.3 |
| 2 | Claude Sonnet 4.5 | 66.9 |
| 3 | GPT-5 | 66.5 |
| 4 | O3 | 65.4 |
| 5 | GLM-4.5V | 64.8 |
| 6 | Qwen 3 VL 32B Instruct | 61.3 |
| 7 | Qwen 3 VL 8B Instruct | 57.1 |
| 8 | Step-3 | 52.8 |
Interactive version: theaggregate.ai/benchmark?slug=autogui-v2-element-captioning · How It Works · Data refreshed daily, snapshot 2026-10-07.