HELM MMLU - Conceptual Physics: leaderboard

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

Top models

#ModelScore
1Gemini 1.5 Pro (001)94.89
2Gemini 1.5 Pro (002)94.47
3DeepSeek V394.04
4GPT-4o (2024-08-06)92.34
5GPT-4o (2024-05-13)91.06
6Claude 3.5 Sonnet (20241022)90.64
7GPT-4 Turbo89.36
8GPT-4 Turbo (Preview)89.36
9Qwen 2.5 72B Instruct88.51
10Claude 3.5 Sonnet (20240620)88.51
11Claude 3 Opus (20240229)88.09
12Llama 3.1 405B Instruct87.66
13Qwen 2 72B Instruct87.23
14GPT-4 (0613)86.81
15Mistral Large 2 (Jul)86.38

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