SmolLM-135M running in pure SQLite at ~40 seconds per token

cneuralnetwork · x · 2026-11-07

A hacker runs SmolLM-135M entirely in SQLite SQL, achieving full inference at roughly 40 seconds per token — an extreme engineering stunt. See the companion GitHub repo post for details.

Related event: Developer Runs SmolLM-135M Entirely Inside SQLite Using Pure SQL(2 posts)→

Original post →

More from Fun

Fun channel →