Science
GPQA Diamond
Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better.
Rankings
Higher is better#196%
GPT-6 Astra
294.4%
GPT-5.4-Pro
394.3%
Gemini 3.1 Pro
494.2%
Claude Opus 4.7
593.6%
GPT-5.5
693.5%
Kimi K3
792.8%
GPT-5.4
892.4%
GPT-5.2
892.4%
Qwen3.7-Max
1091.9%
Gemini 3.0 Pro
1191.7%
Qwen3.8-Flash-NextOW
1291.3%
Claude Opus 4.6
1391.2%
GLM-5.2
1490.9%
DeepSeek-V4.1-Flash
1590.5%
Kimi K2.6
1690.4%
Gemini 3.0 Flash
1790.1%
DeepSeek-V4-Pro
1889.9%
Claude Sonnet 4.6
1989.5%
Muse Spark
2089.2%
Qwen3.8-27BOW
2188.4%
Qwen3.5OW
2288.1%
GPT-5.1
2288.1%
DeepSeek-V4-Flash
2487.6%
Kimi K2.5
2587%
Claude Opus 4.5
2686.7%
Nemotron 3 UltraOS
2786.4%
Gemini 2.5 Pro
2886.2%
GLM-5.1
2986%
GLM-5
2986%
Qwen3.6OW
3185.7%
GLM-4.7
3283.5%
Muse GlimmerOW
3383.4%
Claude Sonnet 4.5
3480.9%
Claude Opus 4.1
3580.1%
gpt-oss-120bOW
3679.6%
Claude Opus 4
3779.2%
Nemotron 3 SuperOS
3879.1%
GLM-4.5
3976%
Llama Nemotron Ultra 253BOW
4075.7%
o1
4175.4%
Claude Sonnet 4
4175.4%
Nemotron 3.5 LightningOS
4375.1%
Kimi K2
4375.1%
Kimi K2 (0905)
4573.3%
o1-preview
4673%
Claude Haiku 4.5
4673%
Nemotron 3 NanoOS
4871.5%
gpt-oss-20bOW
4968%
Claude 3.7 Sonnet
5064%
Nemotron Nano 2OW
5160%
o1-mini
5259.4%
Claude 3.5 Sonnet
5350.4%
Claude 3 Opus
5449.9%
GPT-4o
5542.5%
GPT-4 Turbo
5641.6%
Claude 3.5 Haiku
5740.4%
Claude 3 Sonnet
5840.2%
GPT-4o mini
5933.3%
Claude 3 Haiku
GPQA Diamond — frequently asked questions
Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better.