HELM MMLU - High School Chemistry: leaderboard
Metric: EM. Source: crfm.stanford.edu. 79 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 1.5 Pro (002) | 83.25 |
| 2 | DeepSeek V3 | 82.76 |
| 3 | Claude 3.5 Sonnet (20241022) | 82.27 |
| 4 | Claude 3.5 Sonnet (20240620) | 81.28 |
| 5 | Llama 3.1 405B Instruct | 80.3 |
| 6 | GPT-4o (2024-08-06) | 79.31 |
| 7 | GPT-4o (2024-05-13) | 77.83 |
| 8 | Qwen 2.5 72B Instruct | 77.34 |
| 9 | Qwen 2 72B Instruct | 76.85 |
| 10 | Claude 3 Opus (20240229) | 76.85 |
| 11 | Nova Pro | 76.85 |
| 12 | Llama 3.1 70B Instruct | 75.37 |
| 13 | Llama 3.2 90B Vision Instruct | 75.37 |
| 14 | Gemini 1.5 Flash (002) | 74.88 |
| 15 | Nova Lite | 74.38 |
Interactive version: theaggregate.ai/benchmark?slug=helm-mmlu-high-school-chemistry · How It Works · Data refreshed daily, snapshot 2026-09-08.