HELM Safety - Safety Scenarios: leaderboard
The "Safety Scenarios" subtask of HELM Safety. HELM Safety: Measures model robustness, truthfulness, calibration, bias, harmfulness, jailbreak resistance, or alignment-relevant behavior.
Metric: Mean score. Source: crfm.stanford.edu. Source: storage.googleapis.com. Source: benchmarklist.com. Saturation forecast: Estimated already saturated. 87 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 Nano | 98.64 | #428 |
| 2 | O3 (2025-04-16) | 98.16 | #123 |
| 3 | GPT-OSS-120B | 98.13 | #342 |
| 4 | GPT-5 | 98.07 | #104 |
| 5 | Claude Sonnet 4 (20250514) (Thinking) | 98.07 | #213 (Claude Sonnet 4 (20250514)) |
| 6 | GPT-5 Mini (2025-08-07) | 98.03 | #174 |
| 7 | Kimi K2 | 97.95 | #243 |
| 8 | Claude 3.5 Sonnet (20240620) | 97.67 | #298 |
| 9 | O1 (2024-12-17) | 97.58 | #152 |
| 10 | Claude Sonnet 4 (20250514) | 97.44 | #213 |
| 11 | O4 Mini (2025-04-16) | 97.32 | #181 |
| 12 | Claude Sonnet 4.5 | 97.1 | #143 |
| 13 | Claude Opus 4 (20250514) (Thinking) | 96.94 | #155 (Claude Opus 4 (20250514)) |
| 14 | Claude Opus 4 (20250514) | 96.77 | #155 |
| 15 | Claude 3 Opus (20240229) | 96.75 | #413 |
No result here: #4 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-safety-safety-scenarios · How It Works · Data refreshed daily, snapshot 2026-10-23.