🧠

GPQA Diamond

Graduate-Level Google-Proof Q&A

198 hardest questions in biology, physics, chemistry; PhD experts score 65%

🧠 Knowledge & Reasoning

Best: 94.3%

23 Models Evaluated

Champion

Claude Opus 4.7

Anthropic

94.2%

Best Score

All models ranked by score on GPQA Diamond. Scores are relative to the best score of 94.3%.

#	Model	Provider	Score	vs Best
1	Claude Opus 4.7 Elo 1500	Anthropic	94.2%	Best
2	Claude Opus 4.8	Anthropic	93.6%	-0.7%
3	GPT-5.5 Elo 1478	OpenAI	93.6%	-0.7%
4	GPT-5.4 Pro Elo 1480	OpenAI	93%	-1.3%
5	GPT-5.4 Elo 1463	OpenAI	92.8%	-1.5%
6	Gemini 3.1 Pro Elo 1492	Google	91.9%	-2.4%
7	Claude Opus 4.6 Elo 1503	Anthropic	91.3%	-3.0%
8	DeepSeek V4 Pro (Max)	DeepSeek	90.1%	-4.2%
9	Claude Sonnet 4.6 Elo 1460	Anthropic	89.9%	-4.4%
10	Qwen 3.5 Elo 1450	Alibaba	88.4%	-5.9%
11	Kimi K2.5 Elo 1438	Moonshot	87.6%	-6.7%
12	Kimi K2 Thinking	Moonshot	87.6%	-6.7%
13	Gemini 3 Pro Elo 1486	Google	87%	-7.3%
14	GLM-5 Elo 1454	Zhipu AI	86%	-8.3%
15	MiniMax M2.5 Elo 1404	MiniMax	85.2%	-9.1%
16	Grok 3 Elo 1412	xAI	84.6%	-9.7%
17	MiMo-V2-Flash Elo 1393	Xiaomi	83.7%	-10.6%
18	GPT-oss 120B Elo 1355	OpenAI	80.9%	-13.4%
19	DeepSeek V3.2 Elo 1423	DeepSeek	79.9%	-14.4%
20	Nemotron Ultra 253B Elo 1348	NVIDIA	76%	-18.3%
21	DeepSeek R1 Elo 1398	DeepSeek	71.5%	-22.8%
22	Llama 4 Maverick Elo 1328	Meta	69.8%	-24.5%
23	Mistral Large Elo 1416	Mistral	43.9%	-50.4%

Claude Opus 4.7

94.2%

Claude Opus 4.8

93.6%

GPT-5.5

93.6%

GPT-5.4 Pro

93%

GPT-5.4

92.8%

Gemini 3.1 Pro

91.9%

Claude Opus 4.6

91.3%

DeepSeek V4 Pro (Max)

90.1%

Claude Sonnet 4.6

89.9%

Qwen 3.5

88.4%

Learn how to evaluate your own models on GPQA Diamond.

GPQA Diamond is a standardized benchmark for evaluating AI model capabilities.

To run this benchmark on your own models, visit the official benchmark page or leaderboard for detailed instructions, code, and evaluation scripts.

Data compiled from official model cards, public leaderboards, and independent evaluations.