GPT-5.6 Sol

39.5± 1.0 (95% CI)

Codex · pass rate 84.1% · simplicity 52.9 · $2.14 per library

Runs

Each cell is one full run: the model designs the library, then 3 implementers solve every problem with it. Click a cell to open that run.

TaskRun 1Run 2Run 3Mean ± std

arazu 31.4 35.0 25.230.5 ± 5.0

clirs 49.9 52.5 48.850.4 ± 1.9

heretical 38.2 36.9 36.837.3 ± 0.8

npr 47.0 40.5 42.843.4 ± 3.3

plaid 44.3 44.0 42.043.4 ± 1.2

pyda 45.0 40.9 44.543.5 ± 2.2

canon 55.8 54.3 56.955.6 ± 1.3

roadkill 10.0 8.0 9.89.3 ± 1.1

rubberband 45.5 50.7 52.649.6 ± 3.7

weft 33.2 31.4 36.533.7 ± 2.6

sapi 36.8 38.8 33.436.4 ± 2.7

simu 40.6 43.8 44.342.9 ± 2.0

uglypie 56.0 56.4 45.752.7 ± 6.0

umami 48.5 47.5 48.148.0 ± 0.5

wapi 13.7 21.7 10.515.3 ± 5.7