Vals AI

GPQA Diamond

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

In index
Knowledge
Best in knowledge
Source ↗

Models134

Top result95.5%Gemini 3.1 Pro Preview

Top-3 spread0.8 pts

Updated1 Sep 2026

Result and cost

132 of 135 results

Results

135 results

#ModelResultCostTime

01Gemini 3.1 Pro Preview95.5% ±1.0$0.08166 s

02GPT-5.6 Sol95.2% ±1.1$0.1158 s

03Grok 4.694.7% ±1.1$0.0453 min

04Gemini 3.8 Flash94.4% ±1.5$0.05620 s

05Gemini 3.7 Flash93.9% ±1.5$0.0238.8 s

06Qwen3.8 Max93.7% ±1.2$0.0774 min

07Claude Fable 5.193.4% ±1.9$0.20546 s

07Claude Opus 593.4% ±1.2$0.09540 s

07Gemini 3.6 Flash93.4% ±1.3$0.03218 s

10Claude Fable 593.2% ±1.9$0.452 min

10GPT-5.593.2% ±1.3$0.0832 min

12Kimi K392.9% ±1.3$0.0882 min

12Grok 4.592.9% ±1.3$0.0444 min

14Gemini 3.5 Flash92.7% ±1.5$0.06428 s

14MiniMax M392.7% ±1.4$0.0265 min

16Claude Opus 4.892.4% ±2.0$0.3392 min

16DeepSeek V4 Pro 081392.4% ±2.0$0.0665 min

18GPT-5.6 Luna91.7% ±1.7$0.01154 s

18GPT-5.491.7% ±1.9$0.2093 min

18GPT-5.291.7% ±1.8$0.1493 min

18Gemini 3 Pro91.7% ±1.4$0.0961 s

22Grok 4.391.4% ±1.4$0.02512 min

23Muse Spark 1.191.2% ±2.1$0.03475 s

24GPT-5.6 Terra90.9% ±1.9$0.03437 s

25Claude Opus 4.790.2% ±1.5$0.3653 min

25Qwen3.7 Max90.2% ±1.5$0.0612 min

27DeepSeek V4 Flash 073189.9% ±1.7$0.01380 s

28Claude Opus 4.6 (Adaptive)89.6% ±1.2$0.2955 min

28Muse Spark89.6% ±1.5$0.00312 min

30DeepSeek V4 Pro 042389.4% ±1.7$0.12615 min

31Kimi K2.689.1% ±2.0$0.089 min

32Claude Sonnet 588.9% ±2.2$0.07863 s

32Qwen3.8-27B88.9% ±1.6$0.0472 min

34Grok 4.20 (Reasoning)88.6% ±1.6$0.03459 s

35GLM-5.388.1% ±1.9$0.0564 min

35Grok 488.1% ±1.6$0.1482 min

37Gemini 3 Flash Preview87.9% ±1.6$0.04484 s

38Qwen3.6 Plus87.4% ±1.7$0.0232 min

38Qwen3.5 Plus Thinking87.4% ±1.7$0.0192 min

40Inkling87.1% ±1.8$0.1665 min

41GPT-5.186.6% ±1.7$0.07586 s

41MiniMax M2.786.6% ±2.0$0.0143 min

43GLM-5.3-Flash86.4% ±2.1$0.00162 min

44Claude Opus 4.5 Thinking85.9% ±2.5$0.7132 min

45GLM-5.285.6% ±1.8$0.0574 min

45Claude Sonnet 4.685.6% ±2.4$0.4096 min

45GPT-585.6% ±1.8$0.0733 min

48Grok 4 Fast (Reasoning)85.4% ±1.8$0.00321 s

49Ling 3.0 Flash84.8% ±2.2$0.003344 s

50GLM-5.184.5% ±1.8$0.0454 min

51Grok 4.1 Fast (Reasoning)84.3% ±1.9$0.002968 s

52Kimi K2.5 Thinking84.1% ±2.1$0.033 min

52o384.1% ±1.9$0.03666 s

54Gemini 3.5 Flash-Lite83.8% ±1.8$0.01214 s

55Inkling-Small83.6% ±2.0$0.024 min

56GLM-5 Thinking83.3% ±1.9$0.0344 min

57GPT-5.4 mini83.1% ±2.5$0.01363 s

58Qwen3.5 Flash82.8% ±1.9$0.003480 s

59MiMo-V2.5-Pro82.6% ±2.5$0.00674 min

60MiniMax M2.582.1% ±2.3$0.0183 min

61Claude Sonnet 4.5 Thinking81.6% ±2.3$0.1142 min

61MiMo-V2.581.6% ±2.1$0.002589 s

61Gemini 2.5 Flash (Preview)81.6% ±2.0$0.002919 s

64Gemini 3.1 Flash Lite Preview81.1% ±2.0$0.002316 s

65Gemini 2.5 Pro Exp 03.2580.8% ±2.0$0.01341 s

66DeepSeek V3.2 (Thinking)80.3% ±2.0$0.00354 min

66GPT-5 mini80.3% ±2.0$0.01467 s

68GLM-4.780.0% ±2.0$0.0254 min

69Claude Opus 4.579.5% ±2.5$0.07514 s

69Qwen3 Maxqwen3-max79.5% ±2.0$0.01670 s

71Grok 3 Mini Fast High Reasoning79.3% ±2.0$0.003941 s

72MiniMax M2.178.5% ±2.1$0.0112 min

72Kimi K2 Thinking78.5% ±2.2$0.02310 min

72GPT-OSS 120B78.5% ±2.3$0.006256 s

75Qwen3 Maxqwen3-max-preview77.8% ±2.1$0.0132 min

76GPT-5.4 nano77.5% ±2.4$0.001629 s

77Gemini 2.5 Flash Preview Thinking76.5% ±2.2$0.002824 s

78DeepSeek V3.276.3% ±2.2$0.0006547 s

78Claude Opus 4.1 Thinking76.3% ±2.2$0.14536 s

80o3-mini75.5% ±2.2$0.0362 min

81Claude 3.7 Sonnet Thinking75.3% ±2.2$0.1893 min

82Claude Sonnet 4 Thinking75.0% ±2.2$0.1172 min

83o4-mini74.5% ±2.2$0.01330 s

83GLM-4.674.5% ±2.2$0.0158 min

85Grok 3 [Beta]74.2% ±2.2$0.03930 s

86o173.2% ±2.2$0.25340 s

87Grok 3 Mini Fast Low Reasoning73.0% ±2.2$0.001614 s

88GLM-4.572.2% ±2.3$0.01912 min

88Claude Haiku 4.5 Thinking72.2% ±2.5$0.05583 s

90Claude Opus 471.7% ±2.3$0.06719 s

91Kimi K2 Instruct71.5% ±2.3$0.004541 s

92Qwen3 235B-A22B70.2% ±2.4$0.00655 min

92Gemini 2.5 Flash Lite Preview Thinking70.2% ±2.3$0.0004821 s

94Claude Opus 4.170.0% ±2.3$0.06736 s

95Claude Sonnet 469.4% ±2.3$0.01413 s

95Llama4 Maverick Instruct Basic69.4% ±2.4$0.0009610 s

97GPT-OSS 20B68.9% ±2.5$0.00672 min

98Mistral Large 368.4% ±2.4$0.002221 s

99GPT-4.1 mini67.9% ±2.4$0.00239.3 s

100Claude 3.7 Sonnet67.4% ±2.4$0.0129.2 s

101GPT-4.165.4% ±2.4$0.01423 s

102Grok 4.1 Fast Non Reasoning65.2% ±2.4$0.000458.7 s

102Gemini 2.0 Flash 00165.2% ±2.4$0.000495.8 s

104Gemini 2.5 Flash Lite Preview64.1% ±2.4$0.00127.3 s

105GPT-5 nano63.4% ±2.5$0.004650 s

106Magistral Medium62.4% ±2.4$0.0362 min

106Llama 3.3 Nemotron Super 49B V1 42e84561 Thinking62.4% ±2.5—78 s

108Grok 4 Fast Non Reasoning62.1% ±2.4$0.00113.7 s

109DeepSeek V3 032461.6% ±2.4$0.001623 s

110MiMo-V2-Flash59.3% ±2.5$0.001221 s

110Claude 3.5 Sonnet59.3% ±2.5$0.00917.2 s

112Magistral Small58.3% ±2.5$0.007146 s

112Gemini 1.5 Pro 00258.3% ±2.5$0.00377.7 s

114Gemini 2.5 Flash Preview 04.1757.6% ±2.4$0.0343 s

115Laguna XS.255.0% ±2.5$0.001469 s

116DeepSeek V354.5% ±2.5$0.001521 s

117GPT-4o53.8% ±2.5$0.009515 s

118Grok 250.8% ±2.5$0.009520 s

118GPT-4.1 nano50.8% ±2.5$0.000494.2 s

118Mistral Small 240250.8% ±2.5$0.000598.0 s

121Llama 3.3 70B Instruct Turbo50.0% ±2.5$0.00167.3 s

122Command A48.5% ±2.3$0.0111 s

123Mistral Large 241147.7% ±2.5$0.006716 s

123Llama 3.3 Nemotron Super 49B V1 42e8456147.7% ±2.5—16 s

125Llama 4 Scout 17B 16E Instruct47.0% ±2.5$0.0006512 s

126Gemini 1.5 Flash 00246.0% ±2.5$0.000233.4 s

127GPT-4o mini44.2% ±2.5$0.0005415 s

127Mistral Small 3.144.2% ±2.5$0.000288.8 s

129Claude 3.5 Haiku37.9% ±2.4$0.00226.6 s

130Jamba Large 1.636.1% ±2.1$0.006715 s

131Mistral Medium 3.534.8% ±2.4$0.0954 min

132Jamba Mini 1.632.1% ±2.0$0.000495.0 s

133Command R Plus31.1% ±2.3$0.00748.2 s

134GPT-3.5 Turbo30.6% ±2.3$0.000862.6 s

135Laguna M.127.0% ±1.9—5 min