Introducing talkie: a 13B vintage language model from 1930

Introducing talkie: a 13B vintage language model trained on 260B tokens of pre-1931 English and released under Apache 2.0. The chat variant is finetuned on instruction-response data from pre-1931 texts, with exploration of whether such models can predict the future, invent beyond their cutoffs, or even write code, while striving to avoid post-1931 contamination and pursuing a bootstrapped era-appropriate post-training pipeline.

Source

Simon Willison

Ingéré

28 AVR. 2026 · 08:40

Score édito

3.0 / 5