UbuntuGuard - English: leaderboard
Metric: F1 (%) for PASS/FAIL policy-compliance classification of user-agent dialogues, English policies and dialogues. Source: arxiv.org. Saturation forecast: Estimated already saturated. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama 3.3 70B Instruct | 98.51 |
| 2 | GPT-OSS-20B | 98.48 |
| 3 | DeepSeek V3.1 | 98.42 |
| 4 | gpt-oss-safeguard-20B | 97.26 |
| 5 | GPT-OSS-120B | 94.24 |
Interactive version: theaggregate.ai/benchmark?slug=ubuntuguard-english · How It Works · Data refreshed daily, snapshot 2026-09-26.