Command Palette
Search for a command to run...

Researchers Launch Talkie, a 13B Historical Open-Weight LLM

aiai-modeling 5 posts · 5 accounts

Researchers announced Talkie, an open-weight historical large language model that they said was trained and fine-tuned as a 13 billion-parameter model on a newly curated dataset containing only pre-1930 data. The team described it as a way to interact with a model shaped by language from the past.

One early experiment with the model asked whether a system trained only on older text could learn to code.

From the sources (5 posts)

@iruletheworldmo

alec is a genius and this is incredible, take a look

@soumithchintala

Test of Time LLM from some legends...pretty fun!

@danshipper

Codex is doing AI research for me on Talkie using its in-app browser research question: can an LLM trained only on 1930s text learn to code?

@sleepinyourhat

RT @DavidDuvenaud: Announcing Talkie: a new, open-weight historical LLM! We trained and finetuned a 13B model on a newly-curated dataset o…

@karpathy

RT @status_effects: New work with @AlecRad and @DavidDuvenaud: Have you ever dreamed of talking to someone from the past? Introducing talk…

Preview built on a synthetic news corpus (16 weeks, Apr–Jul 2026). Impact calls are model reads, not price data.

About Archive