HELM MMLU - College Biology: leaderboard

Metric: EM. Source: crfm.stanford.edu. 79 models tracked.

Top models

#ModelScore
1DeepSeek V397.22
2Claude 3 Opus (20240229)97.22
3Claude 3.5 Sonnet (20241022)96.53
4Gemini 1.5 Pro (002)96.53
5Claude 3.5 Sonnet (20240620)96.53
6Llama 3.1 405B Instruct96.53
7Qwen 2.5 72B Instruct94.44
8GPT-4o (2024-08-06)94.44
9Llama 3 70B94.44
10Nova Pro94.44
11Llama 3.3 70B Instruct93.75
12Llama 3.1 70B Instruct93.75
13Llama 3.2 90B Vision Instruct93.75
14GPT-4o (2024-05-13)93.75
15GPT-4 Turbo93.06

Interactive version: theaggregate.ai/benchmark?slug=helm-mmlu-college-biology · How It Works · Data refreshed daily, snapshot 2026-09-08.