Inference
EN

GPQA Diamond · BenchLM

The hardest subset of GPQA featuring the most challenging graduate-level science questions. Sometimes reported separately from the standard GPQA benchmark.

ModelsModels with a result.84
Top resultBest result on this benchmark.96.0%GPT-6 Astra
Top-3 spreadPoints from first to third.0.5 pts
YearYear of release.—

Result and price

45 of 84 results

Results

84 results
#ModelResultThe score from the source.
01GPT-6 Astra96.0%
02Sakana Fugu-Ultra95.5%
02Sakana Fugu95.5%
04GPT-5.6 Sol94.6%
05Gemini 3.1 Pro94.3%
06Kimi K3.1 Pro94.2%
06Claude Opus 4.7 (Adaptive)94.2%
08GPT-5.593.6%
08Claude Opus 4.893.6%
10Step 5 Preview93.5%
10Kimi K393.5%
12Inkling 2 Small93.0%
13GPT-5.6 Terra92.9%
14GPT-5.492.8%
14Ornith-1.5-397B92.8%
16Gemini 3.5 Flash92.7%
17Qwen3.8 Max92.6%
18Qwen3.7 Max92.4%
19Hy4 preview92.3%
19GPT-5.6 Luna92.3%
21Qwen3.8-Flash-Next91.7%
22GLM-5.291.2%
23Qwen3.8-Omni-Flash91.0%
24DeepSeek V4.1 Flash 101790.9%
25Beam 1.0 501B-A23B90.6%
26Kimi K2.690.5%
27Particle-190.4%
28Beam 1.0 501B-A23B FP890.3%
28Qwen3.7 Plus90.3%
30DeepSeek V4 Pro 081390.1%
31Nemotron 3.5 Super VL 120B A12B NVFP489.9%
31Interfaze Beta89.9%
33Nemotron 3.5 Super VL 120B A12B89.5%
33Muse Spark89.5%
33Inkling-Small89.5%
36Beam 1.0 501B-A23B NVFP489.3%
37Mercury 389.2%
37Qwen3.8-27B89.2%
37Claude Opus 4.689.2%
37Ornith-1.5-35B-A3B89.2%
41Solar Pro 489.0%
42Nemotron 3.5 Bolt 74B-A9B BF1688.8%
43Grok 4.2088.5%
44OrcaCyber-Nano-1.088.3%
45DeepSeek V4 Flash 073188.1%
46Inkling87.9%
47Kimi K2.587.6%
48Hy3 Preview87.2%
49MiniMax M2.787.0%
49Nemotron 3 Ultra87.0%
51Nemotron 3.5 Bolt 74B-A9B86.4%
51Ornith-1.5-9B86.4%
53Solar Open 286.3%
54GLM-5.186.2%
55GLM-586.0%
56Ternary Bonsai 2 27B85.8%
57A.X K285.6%
58Ripple-65B Preview85.2%
59Ling 3.0 Flash85.0%
60MAI-Thinking-184.2%
61Ling 3.0 Flash FP884.0%
62K-EXAONE 2.082.2%
63Mercury 2.579.0%
64Gemma 4 12B78.8%
65Sarvam 105B78.7%
66Spark-X2.6-4B77.1%
67Trinity-Large-Thinking76.3%
68Nemotron 3.5 Lightning 30B A3B NVFP475.6%
69LFM2.5-24B-A2B74.4%
70Nemotron 3 Nano Omni 30B A3B72.2%
71ZAYA1-8B71.0%
72MiniCPM5-2B70.2%
73LongCat-Flash-Lite-Sparse69.5%
74Sarvam 30B66.5%
75Mellum2.1-12B-A2.5B-Thinking64.6%
76LFM3-1.7B63.4%
77Trinity-Large-Preview63.3%
78Mellum2-12B-A2.5B-Thinking57.6%
79ZAYA1-74B-Preview57.3%
80LLaDA2.2-mini44.4%
81Soofi S 30B-A3B43.4%
82Mellum2-12B-A2.5B-Instruct40.9%
83MiniCPM5-1B26.3%
84LFM2.5-230M25.4%