AutoGUI-v2 - Element Grounding (Intent): leaderboard

AutoGUI-v2 (2026) tests functionality understanding of GUI regions and elements on web, desktop and mobile screenshots. This board asks the model to locate a UI element from an action intent; a prediction counts when its point falls inside the ground-truth box. Accuracy in percent; higher is better.

Metric: Element grounding by action intent on the AutoGUI-v2 elements, grounding center accuracy (%): the predicted point falls inside the ground-truth box; native resolution, the authors' tuned prompts; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 19 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Qwen 3 VL 32B Instruct81.9#606
2OpenCUA-72B76.8
3GLM-4.5V74.5#754
4Qwen 3 VL 8B Instruct74.1#1029
5OpenCUA-7B71.7
6InfiGUI-G1-7B71.1
7UI-Venus-Ground-72B70.9
8uitars-1.5-7B70#7905
9OpenCUA-32B68.6
10UI-Venus-Ground-7B58.4
11Gemini 2.5 Pro57#276
12Holo2-8B37.9
13OS-Atlas-Base-7B31.3
14UGround-V1-7B22.3
15Qwen 2 VL 7B Instruct17.1#2573

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=autogui-v2-element-grounding-intent · How It Works · Data refreshed daily, snapshot 2026-11-21.