HELM Safety - Anthropic Red Team: leaderboard
The "Anthropic Red Team" subtask of HELM Safety. HELM Safety: Measures model robustness, truthfulness, calibration, bias, harmfulness, jailbreak resistance, or alignment-relevant behavior.
Metric: LM Evaluated Safety score. Source: crfm.stanford.edu. Source: storage.googleapis.com. Saturation forecast: Not forecast. 87 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3 Haiku (20240307) | 99.96 | #771 |
| 2 | Gemini 1.5 Pro (001) | 99.92 | #444 |
| 3 | Gemini 1.5 Flash (001) | 99.92 | #577 |
| 4 | Qwen 3 Next 80B A3B (Thinking) | 99.9 | #320 (Qwen 3 Next 80B A3B) |
| 5 | Claude 3 Sonnet (20240229) | 99.84 | #677 |
| 6 | Claude 3 Opus (20240229) | 99.79 | #412 |
| 7 | Claude 3.5 Sonnet (20240620) | 99.78 | #299 |
| 8 | GPT-4 Turbo | 99.67 | #404 |
| 9 | Claude 3.7 Sonnet (20250219) | 99.66 | #201 |
| 10 | GPT-OSS-20B | 99.65 | #518 |
| 11 | GPT-4.1 Nano | 99.58 | #731 |
| 12 | Granite 4.0 H Small | 99.58 | #665 |
| 13 | Qwen 2.5 72B Instruct | 99.56 | #454 |
| 14 | GPT-3.5 Turbo (0613) | 99.56 | #1118 |
| 15 | GPT-5 Nano | 99.55 | #425 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-safety-anthropic-red-team · How It Works · Data refreshed daily, snapshot 2026-10-20.