MMLU

Models scored

217

Best score

0.88

Metric

EM

MMLU is an AI evaluation tracked by Epoch AI, with 217 scored model versions on record. The scores here are reported by third parties rather than produced by Epoch AI running the evaluation itself.

Scored models span August 2021 to February 2025. The best score recorded is 0.88, measured as EM. Models come from Amazon, Baichuan, Cerebras Systems, DeepMind, Anthropic, Cohere,Cohere for AI, DeepSeek, Databricks and others.

Full record

Score metric: EM
Models scored: 217
Best score recorded: 0.88
Earliest model scored: Aug 5, 2021
Most recent model scored: Feb 26, 2025
Third-party reported: Yes

Leaderboard

01	gpt-4o-2024-11-20	OpenAI	0.88
02	claude-3-5-sonnet-20241022	Anthropic	0.87
03	DeepSeek-V3	DeepSeek	0.87
04	gemini-1.5-pro-002	Google DeepMind	0.87
05	claude-3-5-sonnet-20240620	Anthropic	0.86
06	gpt-4-0314	OpenAI	0.86
07	Llama-3.3-70B-Instruct	Meta AI	0.86
08	gemini-1.5-pro-001	Google DeepMind	0.86
09	qwen2.5-72b-instruct	Alibaba	0.85
10	Qwen2.5-72B	Alibaba	0.85
11	phi-4	Microsoft Research	0.85
12	claude-3-opus-20240229	Anthropic	0.85
13	Llama-3.1-405B-Instruct	Meta AI	0.84
14	Llama-3.1-405B	Meta AI	0.84
15	gpt-4o-2024-08-06	OpenAI	0.84
16	gpt-4o-2024-05-13	OpenAI	0.84
17	gemini-1.5-pro-001-feb24	Google DeepMind	0.83
18	gpt-4-0613	OpenAI	0.82
19	qwen2-72b-instruct	Alibaba	0.82
20	amazon.nova-pro-v1:0	Amazon	0.82
21	gpt-4o-mini-2024-07-18	OpenAI	0.82
22	gpt-4-turbo-2024-04-09	OpenAI	0.81
23	Llama-3.2-90B-Vision-Instruct	Meta AI	0.8
24	Llama-3.1-70B-Instruct	Meta AI	0.8
25	mistral-large-2407	Mistral AI	0.8
26	qwen2.5-14b-instruct	Alibaba	0.8
27	gemini-2.0-flash-exp	Google DeepMind,Google	0.8
28	gpt-4-turbo	OpenAI	0.8
29	Yi-large	01.AI	0.79
30	Meta-Llama-3-70B-Instruct	Meta AI	0.79
31	Qwen2.5-Coder-32B	Alibaba	0.79
32	claude-2.0	Anthropic	0.79
33	DeepSeek-V2	DeepSeek	0.78
34	Phi-3-medium-128k-instruct	Microsoft	0.78
35	gemini-1.5-flash-001	Google DeepMind	0.78
36	Mixtral-8x22B-v0.1	Mistral AI	0.78
37	gemini-1.5-flash-0514	Google DeepMind	0.78
38	amazon.nova-lite-v1:0	Amazon	0.77
39	claude-1.3	Anthropic	0.77
40	Yi-34B	01.AI	0.76
41	claude-3-sonnet-20240229	Anthropic	0.76
42	gemma-2-27b-it	Google DeepMind	0.76
43	Phi-3-small-8k-instruct	Microsoft	0.76
44	Qwen2.5-Coder-14B		0.75
45	qwen1.5-32B	Alibaba	0.74
46	claude-3-5-haiku-20241022	Anthropic	0.74
47	gemini-1.5-flash-002	Google DeepMind	0.74
48	claude-3-haiku-20240307	Anthropic	0.74
49	Yi-34B-Chat	01.AI	0.73
50	claude-2.1	Anthropic	0.73
51	claude-instant-1.1	Anthropic	0.73

More benchmarks

Epoch Capabilities Index Terminal-Bench GPQA Diamond OTIS Mock AIME 2024-2025 GSM8K ARC-AGI ARC-AGI-2 WeirdML

SourceEpoch AI, 'AI Benchmarking Hub'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/benchmarks. Licensed under CC BY 4.0.

← All benchmarks