UI-Vision - Basic: leaderboard

Accuracy (%) at pointing inside the target element when the query names the element directly, over 1,772 desktop-screenshot queries per model.

Metric: Element grounding accuracy (%), Basic setting, 1,772 queries. Source: raw.githubusercontent.com. Saturation forecast: Around 2029. 12 models tracked. Updated 17 Nov 2026.

Top models

#ModelScoreOverall rank
1Aguvis-7B17.8
2OS-Atlas-Base-7B12.2
3CogAgent-9B-2024122012
4Claude 3.7 Sonnet (20250219)9.48#421
5SeeClick9.42
6ShowUI-2B8.07
7MiniCPM-V-2.67.11#3523
8Qwen 2 VL 7B Instruct3.44#2594
9InternVL2.5-8B2.48#2401
10Qwen 2.5 VL 7B Instruct1.24#1893
11Gemini 2.0 Flash (001)0.45#727
12InternVL2-8B0.11#3359

No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.

Interactive version: theaggregate.ai/benchmark?slug=ui-vision-basic · How It Works · Data refreshed daily, snapshot 2026-11-17.