UI-Venus-2 Report - VenusBench-GD: leaderboard
VenusBench-GD is a GUI grounding benchmark over web, desktop and mobile interfaces, covering basic localization, reasoning-heavy cases and refusal for infeasible instructions; English-instruction micro-average point-in-box accuracy. UI-Venus-2 technical report, keeping only runs the authors performed or reproduced, UI-Venus-2 with reasoning off; percent, higher is better.
Metric: Point-in-Box Accuracy (%). Source: arxiv.org. Saturation forecast: Around February 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.7 Plus | 75.2 | #98 |
| 2 | Seed 2.1 Pro | 73.9 | #73 |
| 3 | Kimi K2.6 | 73.1 | #113 |
| 4 | Qwen 3.6 27B | 67.7 | #182 |
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-venusbench-gd · How It Works · Data refreshed daily, snapshot 2026-11-05.