UI-Venus-2 Report - Spatial-CAPTCHA-Bench: leaderboard
Spatial-CAPTCHA-Bench poses 1,050 image-text CAPTCHA instances across seven task formulations that test reference frames, perspective taking, mental rotation and multi-step spatial visualization. The UI-Venus-2 technical report runs every model on the complete set with strict action scoring; Pass@1 in percent, higher is better.
Metric: Pass@1 (%). Source: arxiv.org. Saturation forecast: Around January 2028. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Seed 2.0 Pro | 43.8 | #122 |
| 2 | Qwen 3.6 27B | 31 | #182 |
| 3 | Kimi K2.6 | 24.8 | #113 |
| 4 | Claude Opus 4.6 | 9.5 | #67 |
| 5 | Qwen 3.5 9B | 4.9 | #391 |
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-spatial-captcha-bench · How It Works · Data refreshed daily, snapshot 2026-11-05.