UI-Venus-2 Report - Odysseys - Perfect Rubric Score: leaderboard
Odysseys has 200 long-horizon browsing tasks that span several live websites, judged rubric item by rubric item with gemini-3.1-flash-lite-preview. This series is the share of tasks with every rubric item satisfied, from Ant Group's UI-Venus-2 technical report, keeping only runs the authors performed or reproduced; percent, higher is better.
Metric: Perfect-Rubric Task Rate (%). Source: arxiv.org. Saturation forecast: Around October 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Seed 2.0 Pro | 30.1 | #122 |
| 2 | Qwen 3.6 27B | 18.5 | #182 |
| 3 | Qwen 3.5 9B | 13.5 | #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-odysseys-perfect-rubric-score · How It Works · Data refreshed daily, snapshot 2026-11-05.