HELM MMLU - College Physics: leaderboard

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

Top models

#ModelScore
1Gemini 1.5 Pro (002)86.27
2DeepSeek V381.37
3Claude 3.5 Sonnet (20241022)77.45
4Gemini 1.5 Pro (001)74.51
5Claude 3 Opus (20240229)71.57
6Claude 3.5 Sonnet (20240620)69.61
7Gemini 2.0 Flash (Preview)69.61
8Llama 3.1 405B Instruct69.61
9Gemini 1.5 Flash (001)69.61
10GPT-4o (2024-08-06)68.63
11GPT-4o (2024-05-13)68.63
12Nova Pro64.71
13Gemini 1.5 Flash (002)63.73
14GPT-4 (0613)62.75
15Nova Lite60.78

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