HELM MMLU - Professional Accounting: leaderboard

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

Top models

#ModelScore
1Claude 3.5 Sonnet (20241022)84.04
2DeepSeek V382.62
3Claude 3.5 Sonnet (20240620)81.56
4GPT-4o (2024-08-06)80.14
5Claude 3 Opus (20240229)79.08
6Gemini 1.5 Pro (002)78.72
7GPT-4o (2024-05-13)76.95
8Gemini 2.0 Flash (Preview)73.4
9Llama 3.1 405B Instruct72.7
10Qwen 2 72B Instruct70.92
11Qwen 2.5 72B Instruct70.21
12GPT-4 (0613)70.21
13GPT-4 Turbo (Preview)68.44
14GPT-4 Turbo68.09
15Nova Pro68.09

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