HELM MMLU - Computer Security: leaderboard
Metric: EM. Source: crfm.stanford.edu. 79 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude 3.5 Sonnet (20240620) | 89 |
| 2 | Claude 3.5 Sonnet (20241022) | 87 |
| 3 | Qwen 2.5 72B Instruct | 86 |
| 4 | DeepSeek V3 | 86 |
| 5 | GPT-4 (0613) | 86 |
| 6 | GPT-4 Turbo (Preview) | 86 |
| 7 | Yi Large (Preview) | 86 |
| 8 | Gemini 1.5 Pro (002) | 85 |
| 9 | GPT-4o Mini (2024-07-18) | 85 |
| 10 | GPT-4o (2024-08-06) | 85 |
| 11 | GPT-4o (2024-05-13) | 85 |
| 12 | Qwen 2 72B Instruct | 85 |
| 13 | Llama 3 70B | 85 |
| 14 | Claude 3 Opus (20240229) | 85 |
| 15 | Gemma 2 27B | 84 |
Interactive version: theaggregate.ai/benchmark?slug=helm-mmlu-computer-security · How It Works · Data refreshed daily, snapshot 2026-09-08.