Developer Runs SmolLM-135M Entirely Inside SQLite Using Pure SQL

A developer implemented full inference of the 135M-parameter SmolLM model entirely in pure SQL inside SQLite, including BPE tokenization, a 30-layer Transformer, and KV cache, though it runs at about 40 seconds per token as an extreme engineering stunt.

2026-11-07 ~ 2026-11-07 · 2 related posts

• SmolLM-135M running in pure SQLite at ~40 seconds per token — cneuralnetwork · 2026-11-07
• SmolLM-135M inference in pure SQLite SQL: tokenizer, transformer, KV cache and all — cneuralnetwork · 2026-11-07