AutoGUI-v2 - Element Grounding (Appearance): 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 its appearance description; a prediction counts when its point falls inside the ground-truth box. Accuracy in percent; higher is better.

Metric: Element grounding by appearance description 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 Instruct80.7#606
2OpenCUA-72B77.1
3GLM-4.5V75.4#754
4OpenCUA-7B72.1
5UI-Venus-Ground-72B70.4
6Qwen 3 VL 8B Instruct69.5#1029
7InfiGUI-G1-7B69.4
8uitars-1.5-7B68.9#7905
9OpenCUA-32B67.9
10Gemini 2.5 Pro55.5#276
11UI-Venus-Ground-7B54.6
12Holo2-8B36.3
13OS-Atlas-Base-7B28.7
14UGround-V1-7B18.9
15Qwen 2 VL 7B Instruct15#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-appearance · How It Works · Data refreshed daily, snapshot 2026-11-21.