Dev jokes about running 10 parallel Sol 6.1 sessions to beat seconds-per-token speeds

mgostIH · x · 2026-10-01

Developer mgostIH jokes that Sol 6.1 inference runs at "seconds per token" rather than "tokens per second," forcing him to run 10 parallel sessions just to get work done — a relatable meme about painfully slow model serving.

Original post →

More from Fun

Fun channel →