News for hackers. Stuff that matters.
Re:Inkling 2 Mini by samstevens on Friday November 20, 2026 @11:49PM (#50586044)
when considering gpu hours, larger models are easier/cheaper to train than smaller models. maybe if datacenter inference is infeasible, then it will convince model providers to train small models, but from a pure training perspective, smaller data centers lead to larger models.

There are no comments on this comment yet.