HELM MMLU - High School World History: leaderboard

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

Top models

#ModelScore
1Claude 3.5 Sonnet (20241022)95.78
2GPT-4 Turbo (Preview)95.78
3Claude 3.5 Sonnet (20240620)95.36
4GPT-4o (2024-05-13)94.51
5GPT-4 (0613)94.51
6Llama 3.2 90B Vision Instruct94.09
7GPT-4o (2024-08-06)94.09
8GPT-4 Turbo94.09
9Llama 3.1 405B Instruct94.09
10Llama 3 70B94.09
11Claude 3 Opus (20240229)94.09
12Gemini 1.5 Pro (002)93.67
13Llama 3.1 70B Instruct93.67
14Qwen 2 72B Instruct93.25
15DeepSeek V392.83

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