HELM MMLU - Formal Logic: leaderboard
Metric: EM. Source: crfm.stanford.edu. 79 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude 3.5 Sonnet (20241022) | 78.57 |
| 2 | DeepSeek V3 | 76.98 |
| 3 | Gemini 1.5 Pro (002) | 75.4 |
| 4 | Qwen 2.5 72B Instruct | 73.02 |
| 5 | Nova Pro | 71.43 |
| 6 | GPT-4 Turbo | 70.63 |
| 7 | Gemini 1.5 Pro (001) | 70.63 |
| 8 | Claude 3.5 Sonnet (20240620) | 69.84 |
| 9 | Llama 3.1 405B Instruct | 69.84 |
| 10 | Claude 3 Opus (20240229) | 69.84 |
| 11 | Llama 3.2 90B Vision Instruct | 68.25 |
| 12 | GPT-4o (2024-05-13) | 68.25 |
| 13 | Llama 3.3 70B Instruct | 67.46 |
| 14 | Llama 3.1 70B Instruct | 67.46 |
| 15 | GPT-4o (2024-08-06) | 67.46 |
Interactive version: theaggregate.ai/benchmark?slug=helm-mmlu-formal-logic · How It Works · Data refreshed daily, snapshot 2026-09-08.