Skip to main content
TokenCost logoTokenCost

The Best LLMs on GPQA Diamond

Which model answers the most graduate-level science questions?

103 of the 144 buyable models publish a GPQA Diamond score. It measures graduate-level reasoning, not coding, so it is reported here on its own and nowhere folded into a composite.

Method: Published GPQA Diamond score, descending, exactly as reported. GPQA Diamond is a graduate-level science question set, so it measures reasoning rather than software work, and it is never mixed into the coding quality score. Pricing as of July 2026. 33 rows below carry a score measured at a non-default reasoning tier and are marked as such; every affected model is named on the hub. Read the full method.

RankModelGPQA DiamondAA indexBlended $/1MContext
#1Kimi K3Moonshot93.557.1$6.001.05M
#2Grok 4.5xAI93.153.8$3.00500K
#3MiniMax M3MiniMax92.944.4$1.051M
#4Gemini 3.6 FlashGoogle92.850.1$3.001.05M
#5GPT-5.5OpenAI92.650.4$11.251.05M
#6GPT-5.6 SolOpenAI92.653.6$11.251.05M
#7Claude Fable 5Anthropic92.659.9$20.001M
#8Qwen3.7 MaxAlibaba92.346.0$3.751M
#9Gemini 3.5 FlashGoogle92.145.4$3.381.05M
#10Claude Opus 4.8Anthropic92.055.7$10.001M
#11GPT-5.3 CodexOpenAI91.544.3$4.81400K
#12Claude Opus 4.7Anthropic· non-default tier91.453.5$10.001M
#13Kimi K2.6Moonshot91.144.2$1.71262K
#14Grok 4.20xAI· non-default tier91.137.0$3.002M
#15Claude Sonnet 5Anthropic· non-default tier91.153.4$4.001M
#16DeepSeek V4-ProDeepSeek· non-default tier90.543.1$0.5441M
#17Qwen3.7 PlusAlibaba90.039.0$0.7001M
#18Gemini 3 Flash ReasoningGoogle· non-default tier89.837.8$1.131.05M
#19Muse Spark 1.1Meta89.850.6$2.001M
#20Kimi K2.7 CodeMoonshot89.641.9$1.71262K
#21Claude Opus 4.6Anthropic· non-default tier89.643.7$10.001M
#22GLM-5.2Zhipu· non-default tier89.551.1$2.151M
#23Nex-N2-ProNex AGI89.241.0$1.00262K
#24Grok 4.3xAI89.036.0$1.561M
#25Qwen3.6-Max-PreviewAlibaba88.840.0$2.92262K
#26Qwen3.6-PlusAlibaba88.239.6$0.6201M
#27Kimi K2.5Moonshot· non-default tier87.935.4$1.20262K
#28Grok 4xAI87.733.3$6.002M
#29Claude Sonnet 4.6Anthropic· non-default tier87.547.2$6.001M
#30MiniMax M2.7MiniMax87.438.1$0.525205K
#31GPT-5.1OpenAI· non-default tier87.336.9$3.44400K
#32GPT-5.6 TerraOpenAI87.245.6$5.631.05M
#33GPT-5.4OpenAI· non-default tier87.139.1$5.631.05M
#34MiMo-V2-ProXiaomi87.040.3$1.501.05M
#35GLM-5.1Zhipu· non-default tier86.840.2$2.15200K
#36DeepSeek V4-FlashDeepSeek· non-default tier86.737.5$0.1751M
#37MiMo-V2.5-ProXiaomi86.642.2$0.5441.05M
#38Claude Opus 4.5Anthropic· non-default tier86.640.8$10.00200K
#39GPT-5.2OpenAI86.438.0$4.81400K
#40GLM-4.7Zhipu· non-default tier85.933.7$1.00200K
#41GPT-5.6 LunaOpenAI85.938.1$2.251.05M
#42Qwen 3.5 27BAlibaba· non-default tier85.833.8$0.825128K
#43Gemma 4 31BGoogle· non-default tier85.729.4$0.180262K
#44Grok 4.1 FastxAI· non-default tier85.330.6$0.2752M
#45Grok 4.1 Fast ReasoningxAI· non-default tier85.330.6$0.2752M
#46MiniMax M2.5MiniMax84.833.7$0.525128K
#47GLM-5 TurboZhipu84.738.1$1.90200K
#48o3-proOpenAI84.532.5$35.00200K
#49Gemini 2.5 ProGoogle84.425.8$3.441.05M
#50GPT-5OpenAI84.233.7$3.44400K
#51GPT-5 MediumOpenAI84.233.7$3.44400K
#52Gemini 3.5 Flash-LiteGoogle83.836.5$0.8501.05M
#53Kimi K2 ThinkingMoonshot83.832.7$1.07262K
#54o3OpenAI82.730.4$3.50200K
#55Qwen3.5-Omni PlusAlibaba82.630.6$1.50262K
#56GPT-5.4 MiniOpenAI82.329.8$1.69400K
#57Gemini 3.1 Flash-LiteGoogle82.225.0$0.5631.05M
#58GLM-5Zhipu· non-default tier82.039.5$1.55128K
#59DeepSeek R1DeepSeek81.320.1$0.960128K
#60Step 3.7 FlashStepFun80.930.3$0.438262K
#61Qwen3.5-9BAlibaba· non-default tier80.621.4$0.075262K
#62GPT-5 MiniOpenAI80.330.9$0.688400K
#63Gemma 4 26B A4BGoogle· non-default tier79.225.7$0.120262K
#64Gemini 2.5 FlashGoogle· non-default tier79.020.1$0.8501.05M
#65o4 MiniOpenAI78.425.6$1.93200K
#66LongCat-2.0Meituan78.033.5$1.301M
#67Claude 3.7 SonnetAnthropic· non-default tier77.227.1$6.00200K
#68Mercury 2Inception Labs77.021.4$0.375128K
#69Mistral Small 4Mistral· non-default tier76.919.6$0.263256K
#70Nova 2.0 LiteAmazon76.819.0$0.8501M
#71GPT-5.4 NanoOpenAI76.130.2$0.463400K
#72Command ACohere76.122.5$4.38128K
#73Command A+Cohere76.122.5$4.38128K
#74Gemma 4 12BGoogle· non-default tier75.321.8$0.00262K
#75o3 MiniOpenAI74.819.0$1.93200K
#76Mistral Medium 3.5Mistral74.829.9$3.00256K
#77o1OpenAI74.723.4$26.25200K
#78Qwen3.5-Omni FlashAlibaba74.219.0$0.850262K
#79Qwen3 Coder NextAlibaba73.721.1$0.282262K
#80Grok 3xAI69.318.4$6.00131K
#81Mistral Large 3Mistral68.015.9$0.750262K
#82Magistral MediumMistral67.912.5$2.7540K
#83GPT-OSS 120BOpenAI· non-default tier67.214.9$0.077131K
#84Llama 4 MaverickMeta67.114.3$0.4151.05M
#85GPT-5 NanoOpenAI67.019.0$0.138128K
#86GPT-4.1OpenAI66.619.4$3.501.05M
#87GPT-4.1 MiniOpenAI66.414.8$0.7001.05M
#88Magistral SmallMistral64.110.7$0.75040K
#89Gemini 2.5 Flash-LiteGoogle· non-default tier62.511.4$0.1751.05M
#90GPT-OSS 20BOpenAI· non-default tier61.114.3$0.131131K
#91Llama 4 ScoutMeta58.710.0$0.1351.05M
#92GLM-4.7-flashZhipu· non-default tier58.122.9$0.152200K
#93Sonar ProPerplexity57.89.3$6.00128K
#94Ministral 3 14BMistral57.211.1$0.200262K
#95GPT-4oOpenAI· non-default tier54.311.2$4.38128K
#96GPT-4.1 NanoOpenAI51.29.6$0.1751.05M
#97Grok 2xAI51.08.0$4.00131K
#98Pixtral LargeMistral50.58.1$3.00128K
#99Ministral 3 8BMistral47.19.0$0.150262K
#100GPT-4o MiniOpenAI42.66.9$0.263128K
#101Devstral SmallMistral· non-default tier41.49.3$0.150256K
#102Mistral Small 3.2Mistral· non-default tier38.14.7$0.150128K
#103Mistral MediumMistral34.93.6$0.800131K

Frequently asked

Nearby questions, answered elsewhere

All rankings