OSWorld-G - Text Matching: leaderboard

OSWorld-G samples whose target is located by its visible text (268 of 564); accuracy of the model's click.

Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 6 models tracked. Updated 17 Nov 2026.

Top models

#ModelScoreOverall rank
1Jedi 3B-1080p67.4
2Jedi-7B-1080p65.9
3Qwen 2.5 VL 32B Instruct63.2#1074
4UGround-V1-7B51.3
5Qwen 2.5 VL 7B Instruct45.6#1893
6Qwen 2.5 VL 3B Instruct41.4#2648

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

Interactive version: theaggregate.ai/benchmark?slug=osworld-g-text-matching · How It Works · Data refreshed daily, snapshot 2026-11-17.