Knowledge Benchmarks

General knowledge and factual understanding - Compare AI models across 4 specialized benchmarks including MMLU, ARC-Challenge, HellaSwag, GPQA, and more.

Filters & Search

Filter models by creator, type, reasoning, or search by name to find the perfect AI model for your needs

Knowledge Benchmark Results

Showing 25 of 52 models • Click column headers to sort


1 GPT-5 (high) OpenAI	OpenAI	Proprietary	Reasoning	128K	72	93%	91%	89%	87%
2 o1-preview OpenAI	OpenAI	Proprietary	Reasoning	200K	71	92%	90%	88%	86%
3 GPT-5 (medium) OpenAI	OpenAI	Proprietary	Reasoning	128K	70	91%	89%	87%	85%
4 Grok 4 xAI	xAI	Proprietary	Non-Reasoning	128K	69	87%	86%	84%	82%
5 GPT-5 mini OpenAI	OpenAI	Proprietary	Reasoning	128K	68	88%	86%	84%	82%
6 o3-pro OpenAI	OpenAI	Proprietary	Reasoning	200K	68	88%	89%	87%	85%
7 o3 OpenAI	OpenAI	Proprietary	Reasoning	200K	67	86%	87%	85%	83%
8 Qwen2.5-1M Alibaba	Alibaba	Open Weight	Non-Reasoning	1M	66	84%	83%	81%	79%
9 Qwen2.5-72B Alibaba	Alibaba	Open Weight	Non-Reasoning	128K	65	83%	82%	80%	78%
10 o4-mini (high) OpenAI	OpenAI	Proprietary	Non-Reasoning	200K	65	82%	82%	80%	78%
11 Gemini 2.5 Pro Google	Google	Proprietary	Non-Reasoning	2M	65	83%	83%	81%	79%
12 DeepSeek Coder 2.0 DeepSeek	DeepSeek	Open Weight	Non-Reasoning	128K	64	80%	79%	77%	75%
13 DeepSeek LLM 2.0 DeepSeek	DeepSeek	Open Weight	Non-Reasoning	128K	63	79%	78%	76%	74%
14 Claude 4.1 Opus Anthropic	Anthropic	Proprietary	Non-Reasoning	200K	61	76%	76%	74%	72%
15 Claude 4 Sonnet Anthropic	Anthropic	Proprietary	Non-Reasoning	200K	59	73%	73%	71%	69%
16 Llama 3.1 405B Meta	Meta	Open Weight	Non-Reasoning	128K	58	70%	70%	68%	66%
17 Mistral Large 2 Mistral	Mistral	Proprietary	Non-Reasoning	128K	57	68%	68%	66%	64%
18 GPT-4o OpenAI	OpenAI	Proprietary	Non-Reasoning	128K	56	66%	66%	64%	62%
19 Claude 3.5 Sonnet Anthropic	Anthropic	Proprietary	Non-Reasoning	200K	55	65%	65%	63%	61%
20 Gemini 1.5 Pro Google	Google	Proprietary	Non-Reasoning	2M	54	64%	64%	62%	60%
21 Mistral 8x7B Mistral	Mistral	Open Weight	Non-Reasoning	32K	52	65%	64%	62%	60%
22 Gemini 1.0 Pro Google	Google	Proprietary	Non-Reasoning	32K	52	62%	62%	60%	58%
23 Claude 3 Opus Anthropic	Anthropic	Proprietary	Non-Reasoning	200K	51	61%	61%	59%	57%
24 GPT-4 Turbo OpenAI	OpenAI	Proprietary	Non-Reasoning	128K	50	60%	60%	58%	56%
25 Llama 3 70B Meta	Meta	Open Weight	Non-Reasoning	128K	48	58%	58%	56%	54%

Showing 25 of 52 models

About Knowledge Benchmarks

MMLU

Tests knowledge across 57 academic subjects

GPQA

Expert-level questions in biology, physics, and chemistry

SuperGPQA

Enhanced version covering 285 disciplines

OpenBookQA

Multi-step reasoning with scientific facts