HELM MMLU - Security Studies: leaderboard
The "Security Studies" subtask of HELM MMLU. HELM Classic MMLU group leaderboard from the public HELM release artifacts.
Metric: EM. Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 79 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3 Opus (20240229) | 88.57 | #413 |
| 2 | Claude 3.5 Sonnet (20241022) | 88.16 | #291 |
| 3 | Claude 3.5 Sonnet (20240620) | 87.76 | #298 |
| 4 | Gemini 1.5 Pro (001) | 87.35 | #446 |
| 5 | Mistral Large 2 (Jul) | 86.53 | #485 |
| 6 | Claude 3 Sonnet (20240229) | 86.53 | #681 |
| 7 | Mixtral 8x22B | 86.53 | #770 |
| 8 | GPT-4 (0613) | 86.12 | #494 |
| 9 | Gemini 1.5 Pro (002) | 85.71 | #350 |
| 10 | Llama 3.1 405B Instruct | 85.71 | #461 |
| 11 | Llama 3.2 90B Vision Instruct | 85.31 | #626 |
| 12 | Llama 3.1 70B Instruct | 84.9 | #566 |
| 13 | Qwen 2.5 72B Instruct | 84.9 | #458 |
| 14 | Nova Pro | 84.9 | #573 |
| 15 | Llama 3.3 70B Instruct | 84.49 | #536 |
No result here: #4 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-mmlu-security-studies · How It Works · Data refreshed daily, snapshot 2026-10-23.