Time per Task includes search time and model time. Keenable takes 15.0s (realtime) and 15.5s (pro) per task, 2nd and 3rd fastest on the board, behind Octen (14.6s), against a board median of about 28s. Search takes only 2.3s and 3.0s of that; the rest is model time.
On Artificial Analysis Search Index vs. Time per Task, both modes are on the frontier: no product scores higher in less time.
