SketchBench

Full Table

Detailed per-run data

RankModelSolvedFailedGuessesCostTimeCompleted
#1GPT-5.6 Luna (xhigh)77/1000680$0.233991.0s30/07/2026, 19:50:49
#2GPT-5.4 Mini (high)77/1000695$2.745527.3s18/03/2026, 15:23:50
#3GPT-5.6 Luna (medium)76/1000717$0.081390.5s30/07/2026, 19:53:56
#4Gemini 3 Flash (dynamic)75/1000716$0.28665.5s12/03/2026, 03:08:38
#5GPT-5.6 Luna (high)74/1000754$0.132537.9s30/07/2026, 19:33:52
#6GPT-5 Mini (medium)71/1003855$0.924739.9s12/03/2026, 03:09:20
#7Gemini 3.1 Flash Lite (medium)62/1000978$0.32788.9s12/03/2026, 03:09:27
#8GPT-5.4 Mini (medium)60/10001019$0.841649.0s17/03/2026, 17:51:34
#9Gemini 3.1 Flash Lite (minimal)46/10001294$0.10394.0s12/03/2026, 03:09:31
#10GPT-5.4 Nano (medium)44/10001326$0.251797.2s17/03/2026, 17:53:17
#11Claude Haiku 4.533/10001522$0.43597.4s12/03/2026, 03:09:38