HELM MMLU - Global Facts Leaderboard — The Aggregate

HELM MMLU - Global Facts: leaderboard

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

Top models

#	Model	Score
1	Claude 3.5 Sonnet (20241022)	80
2	Gemini 1.5 Pro (002)	77
3	Claude 3.5 Sonnet (20240620)	72
4	Llama 3.1 405B Instruct	71
5	GPT-4o (2024-08-06)	69
6	DeepSeek V3	68
7	Gemini 2.0 Flash (Preview)	66
8	Claude 3 Opus (20240229)	66
9	Gemini 1.5 Pro (001)	66
10	GPT-4o (2024-05-13)	64
11	GPT-4 (0613)	62
12	Llama 3.1 70B Instruct	61
13	Qwen 2.5 72B Instruct	61
14	Llama 3.2 90B Vision Instruct	60
15	Llama 3.3 70B Instruct	58

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