Unique

Talkie

Language model trained only on text published before 1931

Talkie

A 13B-parameter LLM whose training corpus is exclusively text in the U.S. public domain — anything published before January 1, 1931. Talk to it about chemistry and you’ll get pre-quantum answers; ask its opinion on cinema and “the talkies” are still novel. A genuine artifact of cultural history, and a deliberate experiment in what an LLM looks like when no copyright concerns clouded its training set.

talkie-lm.com ↗

← back