GPQA Diamond

Name: GPQA Diamond
Creator: AIB
License: https://www.gnu.org/licenses/gpl-3.0.html

About This Benchmark

A multiple-choice benchmark measuring graduate-level expert knowledge in biology, chemistry, and physics — challenging even for domain experts. Score is accuracy (%).

Source: Artificial Analysis

Rank	Model
#1	Google Gemini 3.1 Pro	94.1%
#2	OpenAI GPT-5.6 Sol	94.1%
#3	OpenAI GPT-5.5	93.5%
#4	Moonshot AI Kimi K3	93.5%
#5	Anthropic Claude Opus 5	93.2%
#6	xAI Grok 4.5	93.1%
#7	MiniMax MiniMax M3	92.9%
#8	Google Gemini 3.6 Flash	92.8%
#9	Anthropic Claude Fable 5	92.6%
#10	OpenAI GPT-5.6 Terra	92.5%
#11	Alibaba Qwen: Qwen3.7 Max	92.3%
#12	Google Gemini 3.5 Flash	92.2%
#13	Anthropic Claude Opus 4.8	92.0%
#14	OpenAI GPT-5.4	92.0%
#15	Anthropic Claude Opus 4.7	91.4%
#16	Anthropic Claude Sonnet 5	91.1%
#17	OpenAI GPT-5.6 Luna	91.1%
#18	xAI Grok 4.20 (Reasoning)	91.1%
#19	Moonshot AI Kimi K2.6	91.1%
#20	xAI Grok 4.3	90.1%
#21	Alibaba Qwen3.7 Plus	90.0%
#22	Google Gemini 3 Flash	89.8%
#23	Meta Muse Spark 1.1	89.8%
#24	Tencent Hy3	89.7%
#25	Anthropic Claude Opus 4.6	89.6%
#26	Moonshot AI Kimi K2.7 Code	89.6%
#27	Z.ai GLM 5.2	89.5%
#28	DeepSeek DeepSeek V4 Flash	89.4%
#29	DeepSeek DeepSeek V4 Pro	88.8%
#30	Alibaba Qwen3.6 Max	88.8%
#31	Meta Muse Spark	88.4%
#32	Alibaba Qwen3.6 Plus	88.2%
#33	Moonshot AI Kimi K2.5	87.9%
#34	MiniMax MiniMax M2.7	87.4%
#35	Xiaomi MiMo V2 Pro	87.0%
#36	Z.ai GLM-5.1	86.8%
#37	NVIDIA Nemotron 3 Ultra	86.7%
#38	Anthropic Claude Opus 4.5	86.6%
#39	Xiaomi MiMo V2.5	86.6%
#40	Xiaomi MiMo V2.5 Pro	86.6%
#41	Alibaba Qwen3.5 397B A17B	86.1%
#42	Google Gemma 4 31B	85.7%
#43	Ring AI Ring-2.6-1T	85.7%
#44	xAI Grok 4.1 Fast (Reasoning)	85.3%
#45	MiniMax MiniMax M2.5	84.8%
#46	Z.ai GLM-5-Turbo	84.7%
#47	Google Gemini 2.5 Pro	84.4%
#48	Alibaba Qwen: Qwen3.6 35B A3B	84.1%
#49	DeepSeek DeepSeek V3.2	84.0%
#50	Google Gemini 3.5 Flash-Lite	83.8%
#51	Anthropic Claude Sonnet 4.5	83.4%
#52	OpenAI GPT-5.4 Mini	82.3%
#53	Google Gemini 3.1 Flash Lite	82.2%
#54	Z.ai GLM-5	82.0%
#55	Anthropic Claude Opus 4.1	80.9%
#56	Z.ai GLM 5V Turbo	80.9%
#57	Stepfun Step 3.7 Flash	80.9%
#58	NVIDIA Nemotron 3 Super	80.0%
#59	Anthropic Claude Sonnet 4.6	79.7%
#60	Anthropic Claude Opus 4	79.6%
#61	Google Gemini 2.5 Flash	79.0%
#62	LG AI Research K-EXAONE	78.3%
#63	Meituan LongCat 2.0	78.0%
#64	Anthropic Claude Sonnet 4	77.7%
#65	Baidu ERNIE 5.0 Thinking	77.7%
#66	xAI Grok 4.20	77.6%
#67	Mistral AI Mistral Small 4	76.9%
#68	OpenAI GPT-5.4 Nano	76.1%
#69	NVIDIA Nemotron 3 Nano 30B A3B	75.7%
#70	Cohere North Mini Code	75.7%
#71	Arcee AI Trinity Large Thinking	75.2%
#72	Mistral AI Mistral Medium 3.5	74.8%
#73	Upstage Solar Pro 3	72.4%
#74	OpenAI GPT-5 Mini	68.7%
#75	OpenAI GPT-5	68.6%
#76	Anthropic Claude Haiku 4.5	67.2%
#77	OpenAI GPT OSS 120B	67.2%
#78	Meta Llama 4 Maverick	67.1%
#79	OpenAI GPT-5 Nano	67.0%
#80	OpenAI GPT-4.1	66.6%
#81	Google Gemini 2.5 Flash Lite	65.1%
#82	xAI Grok 4.1 Fast	63.7%
#83	Meituan Longcat Flash Chat	63.6%
#84	Amazon Nova 2 Lite	60.3%
#85	Meta Llama 4 Scout	58.7%
#86	Perplexity Sonar	47.1%
#87	NVIDIA Nemotron 3 Nano Omni	46.9%
#88	IBM Granite 4.1 8B	43.3%