HELM Safety Mean score: leaderboard
HELM Safety mean score aggregates safety-oriented HELM scenarios into a single model score.
Metric: Score (%). Source: storage.googleapis.com. 87 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 nano 2025-08-07 | 98.6 | |
| 2 | o3 2025-04-16 | 98.2 | |
| 3 | Claude 4 Sonnet (20250514, extended thinking) | 98.1 | |
| 4 | GPT-5 2025-08-07 | 98.1 | |
| 5 | gpt-oss-120b | 98.1 | |
| 6 | GPT-5 mini 2025-08-07 | 98 | |
| 7 | Kimi K2 Instruct | 97.9 | |
| 8 | Claude 3.5 Sonnet 20240620 | 97.7 | |
| 9 | o1 2024-12-17 | 97.6 | |
| 10 | Claude 4 Sonnet 20250514 | 97.4 |
Interactive version: theaggregate.ai/benchmark?slug=helm-safety-mean-score · How It Works · Data refreshed daily, snapshot 2026-10-20.