UI-Vision - Functional: leaderboard

Accuracy (%) at pointing inside the target element when the query describes what the element does instead of naming it, 1,772 queries per model.

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

Top models

#ModelScoreOverall rank
1Aguvis-7B18.3
2CogAgent-9B-2024122012.2
3OS-Atlas-Base-7B11.2
4Claude 3.7 Sonnet (20250219)7.73#421
5ShowUI-2B7.67
6MiniCPM-V-2.65.3#3523
7SeeClick4.68
8Qwen 2 VL 7B Instruct3.22#2594
9InternVL2.5-8B2.82#2401
10Qwen 2.5 VL 7B Instruct0.79#1893
11Gemini 2.0 Flash (001)0.4#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-functional · How It Works · Data refreshed daily, snapshot 2026-11-17.