HELM Safety Anthropic Red Team: leaderboard
Metric: LM Evaluated Safety score (%). Source: storage.googleapis.com. 87 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude 3 Haiku (20240307) | 100 |
| 2 | Qwen 3 Next 80B A3B (Thinking) | 99.9 |
| 3 | Gemini 1.5 Pro (001) | 99.9 |
| 4 | Gemini 1.5 Flash (001) | 99.9 |
| 5 | Granite 4.0 Small with guardian | 99.9 |
| 6 | Qwen3 235B A22B Instruct 2507 FP8 | 99.9 |
| 7 | Claude 3.5 Sonnet (20240620) | 99.8 |
| 8 | Claude 3 Opus (20240229) | 99.8 |
| 9 | Claude 3 Sonnet (20240229) | 99.8 |
| 10 | GPT-OSS-20B | 99.7 |
| 11 | Claude 3.7 Sonnet (20250219) | 99.7 |
| 12 | GPT-4 Turbo | 99.7 |
| 13 | GPT-4.1 Nano | 99.6 |
| 14 | GPT-5 Nano | 99.6 |
| 15 | GPT-3.5 Turbo (0613) | 99.6 |
Interactive version: theaggregate.ai/benchmark?slug=helm-safety-anthropic-red-team · How It Works · Data refreshed daily, snapshot 2026-09-05.