Vals AI

GPQA Diamond

Graduate-level Google-Proof Q&A benchmark evaluating models on questions that require deep reasoning.

In index
Knowledge
Source ↗

Models128

Top result95.5%Gemini 3.1 Pro Preview

Top-3 spread0.8 pts

Updated1 Sept 2026

Result and cost

Results

129 results

#ModelResultCostTime

Gemini 3.1 Pro Preview95.5%$0.08166 s

GPT-5.6 Sol95.2%$0.1158 s

Grok 4.694.7%$0.0453 min

Gemini 3.8 Flash94.4%$0.05620 s

Gemini 3.7 Flash93.9%$0.0239 s

Qwen3.8 Max93.7%$0.0774 min

Claude Fable 5.193.4%$0.20546 s

Claude Opus 593.4%$0.09540 s

Gemini 3.6 Flash93.4%$0.03218 s

Claude Fable 593.2%$0.452 min

GPT-5.593.2%$0.0832 min

Grok 4.592.9%$0.0444 min

Gemini 3.5 Flash92.7%$0.06428 s

MiniMax M392.7%$0.0265 min

Claude Opus 4.892.4%$0.3392 min

GPT-5.6 Luna91.7%$0.01154 s

GPT-5.491.7%$0.2093 min

GPT-5.291.7%$0.1493 min

Gemini 3 Pro91.7%$0.0961 s

Grok 4.391.4%$0.02512 min

Muse Spark 1.191.2%$0.03475 s

GPT-5.6 Terra90.9%$0.03437 s

DeepSeek V4 Pro 042390.9%$0.0665 min

Claude Opus 4.790.2%$0.3653 min

Qwen3.7 Max90.2%$0.0612 min

DeepSeek V4 Flash 042389.9%$0.01380 s

Claude Opus 4.6 (Adaptive)89.6%$0.2955 min

Muse Spark89.6%<$0.012 min

Claude Sonnet 588.9%$0.07863 s

Qwen3.8-27B88.9%$0.0472 min

Grok 4.20 (Reasoning)88.6%$0.03459 s

GLM-5.388.1%$0.0564 min

Grok 488.1%$0.1482 min

Gemini 3 Flash Preview87.9%$0.04484 s

Qwen3.6 Plus87.4%$0.0232 min

Qwen3.5 Plus Thinking87.4%$0.0192 min

37TMInkling87.1%$0.1665 min

GPT-5.186.6%$0.07586 s

MiniMax M2.786.6%$0.0143 min

GLM-5.3-Flash86.4%<$0.012 min

Claude Opus 4.5 Thinking85.9%$0.7132 min

GLM-5.285.6%$0.0574 min

Claude Sonnet 4.685.6%$0.4096 min

GPT-585.6%$0.0733 min

Grok 4 Fast (Reasoning)85.4%<$0.0121 s

46INLing 3.0 Flash84.8%<$0.0144 s

GLM-5.184.5%$0.0454 min

Grok 4.1 Fast (Reasoning)84.3%<$0.0168 s

49MAKimi K2.5 Thinking84.1%$0.033 min

o384.1%$0.03666 s

Gemini 3.5 Flash-Lite83.8%$0.01214 s

52TMInkling-Small83.6%$0.024 min

GLM-5 Thinking83.3%$0.0344 min

GPT-5.4 mini83.1%$0.01363 s

Qwen3.5 Flash82.8%<$0.0180 s

MiMo-V2.5-Pro82.6%<$0.014 min

Qwen3 Max82.2%$0.0746 min

MiniMax M2.582.1%$0.0183 min

Claude Sonnet 4.5 Thinking81.6%$0.1142 min

MiMo-V2.581.6%<$0.0189 s

Gemini 2.5 Flash (Preview)81.6%<$0.0119 s

Gemini 3.1 Flash Lite Preview81.1%<$0.0116 s

Gemini 2.5 Pro Exp 03.2580.8%$0.01341 s

DeepSeek V3.2 (Thinking)80.3%<$0.014 min

GPT-5 mini80.3%$0.01467 s

GLM-4.780.0%$0.0254 min

Claude Opus 4.579.5%$0.07514 s

Grok 3 Mini Fast High Reasoning79.3%<$0.0141 s

MiniMax M2.178.5%$0.0112 min

69MAKimi K2 Thinking78.5%$0.02310 min

GPT-OSS 120B78.5%<$0.0156 s

Qwen3 Max77.8%$0.0132 min

GPT-5.4 nano77.5%<$0.0129 s

Gemini 2.5 Flash Preview Thinking76.5%<$0.0124 s

DeepSeek V3.276.3%<$0.0147 s

Claude Opus 4.1 Thinking76.3%$0.14536 s

o3-mini75.5%$0.0362 min

Claude 3.7 Sonnet Thinking75.3%$0.1893 min

Claude Sonnet 4 Thinking75.0%$0.1172 min

o4-mini (high)74.5%$0.01330 s

GLM-4.674.5%$0.0158 min

Grok 3 [Beta]74.2%$0.03930 s

o173.2%$0.25340 s

Grok 3 Mini Fast Low Reasoning73.0%<$0.0114 s

GLM-4.572.2%$0.01912 min

Claude Haiku 4.5 Thinking72.2%$0.05583 s

Claude Opus 471.7%$0.06719 s

88MAKimi K2 Instruct71.5%<$0.0141 s

Gemini 2.5 Flash Lite Preview Thinking70.2%<$0.0121 s

Qwen3 235B A22B70.2%<$0.015 min

Claude Opus 4.170.0%$0.06736 s

Claude Sonnet 469.4%$0.01413 s

Llama4 Maverick Instruct Basic69.4%<$0.0110 s

GPT-OSS 20B68.9%<$0.012 min

GPT-4.1 mini67.9%<$0.019 s

Claude 3.7 Sonnet67.4%$0.0129 s

GPT-4.165.4%$0.01423 s

Grok 4.1 Fast Non Reasoning65.2%<$0.019 s

Gemini 2.0 Flash 00165.2%<$0.016 s

100

Gemini 2.5 Flash Lite Preview64.1%<$0.017 s

101

GPT-5 nano63.4%<$0.0150 s

102

Magistral Medium62.4%$0.0362 min

102

Llama 3.3 Nemotron Super 49B V1 42e84561 Thinking62.4%—78 s

104

Grok 4 Fast Non Reasoning62.1%<$0.014 s

105

MiMo-V2-Flash59.3%<$0.0121 s

105

Claude 3.5 Sonnet59.3%<$0.017 s

107

Magistral Small58.3%<$0.0146 s

107

Gemini 1.5 Pro 00258.3%<$0.018 s

109

DeepSeek V358.1%<$0.0123 s

109

Mistral Large58.1%<$0.0121 s

111

Gemini 2.5 Flash Preview 04.1757.6%$0.0343 s

112POLaguna XS.255.0%<$0.0169 s

113

GPT-4o52.0%$0.01510 s

114

GPT-4.1 nano50.8%<$0.014 s

114

Grok 250.8%<$0.0120 s

116

Llama 3.3 70B Instruct Turbo50.0%<$0.017 s

117COCommand A48.5%$0.0111 s

118

Llama 3.3 Nemotron Super 49B V1 42e8456147.7%—16 s

119

Mistral Small47.5%<$0.018 s

120

Llama 4 Scout 17B 16E Instruct47.0%<$0.0112 s

121

Gemini 1.5 Flash 00246.0%<$0.013 s

122

GPT-4o mini44.2%<$0.0115 s

123

Claude 3.5 Haiku37.9%<$0.017 s

124ALJamba Large 1.636.1%<$0.0115 s

125

Mistral Medium 3.534.8%$0.0954 min

126ALJamba Mini 1.632.1%<$0.015 s

127COCommand R Plus31.1%<$0.018 s

128

GPT-3.5 Turbo30.6%<$0.013 s

129POLaguna M.127.0%—5 min