HELM Safety BBQ: leaderboard
Metric: BBQ accuracy (%). Source: storage.googleapis.com. 87 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude 4 Opus 20250514 | 99.3 |
| 2 | Claude 4.5 Sonnet 20250929 | 98.9 |
| 3 | gpt-oss-120b | 98.5 |
| 4 | Gemini 3 Pro Preview | 98.4 |
| 5 | Claude 4 Sonnet 20250514 | 97.9 |
| 6 | o3 2025-04-16 | 97.9 |
| 7 | GLM-4.5-Air-FP8 | 97.8 |
| 8 | Gemini 2.5 Flash 04-17 preview | 97.7 |
| 9 | GPT-5 nano 2025-08-07 | 97.6 |
| 10 | o1 2024-12-17 | 97.3 |
Interactive version: theaggregate.ai/benchmark?slug=helm-safety-bbq-ixk8de · How It Works · Data refreshed daily, snapshot 2026-10-09.