UI-Venus-2 Report - OSWorld-G-R: leaderboard
OSWorld-G-R evaluates grounding of functionally relevant interface elements in complex desktop environments. From Ant Group's UI-Venus-2 technical report, keeping only cells the authors evaluated: two baselines and UI-Venus-2 run with reasoning off. Accuracy in percent, higher is better.
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around February 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.6 27B | 76.9 | #182 |
| 2 | Kimi K2.6 | 69.7 | #113 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=ui-venus-2-report-osworld-g-r · How It Works · Data refreshed daily, snapshot 2026-11-05.