UI-Vision - Spatial: leaderboard

Accuracy (%) at pointing inside the target element when the query locates it relative to another element (left of, below), 1,935 queries per model.

Metric: Element grounding accuracy (%), Spatial setting, 1,935 queries. Source: raw.githubusercontent.com. Saturation forecast: Around August 2028. 10 models tracked. Updated 17 Nov 2026.

Top models

#ModelScoreOverall rank
1Claude 3.7 Sonnet (20250219)7.6#421
2Aguvis-7B5.06
3OS-Atlas-Base-7B3.67
4CogAgent-9B-202412202.63
5ShowUI-2B2.07
6SeeClick2.07
7Qwen 2 VL 7B Instruct1.45#2594
8InternVL2.5-8B0.98#2401
9Qwen 2.5 VL 7B Instruct0.51#1893
10Gemini 2.0 Flash (001)0.05#727

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-spatial · How It Works · Data refreshed daily, snapshot 2026-11-17.