Plain-AyI
← All projects
Training & fine-tuning

nanochat

A small, simple code base for training your own ChatGPT-style chatbot from scratch and chatting with it.

New this week

What it does

nanochat is a compact set of code for training a large language model (LLM, the kind of AI behind chatbots) from start to finish. It covers every main stage: breaking text into pieces the model can read, the main training, extra fine-tuning, testing, and finally chatting with the result.

Its makers say you can train a model about as capable as GPT-2 (a 2019 model that cost far more to train) in a couple of hours on a rented cloud machine with eight powerful graphics cards, for well under $100. You then talk to it in a simple text window. The README notes the result is a bit like talking to a kindergartener.

One setting, the model's depth (its number of layers), controls the size. All the other training settings are worked out automatically. The project also keeps a public leaderboard of who can train a GPT-2-level model fastest, and it includes a smaller version that runs on an ordinary computer or Apple Silicon Mac, though with weak results.

Who it's for

People learning how chatbots are built, and researchers who want a small, easy-to-change setup for experimenting with model training.

Can I use it without coding?

Developers only. You run Python scripts from the command line and usually rent a powerful GPU server, so programming comfort is needed. New to this? Start here.

Can I run it?

Cost
FreeNothing to pay or sign up for.
Runs on
Linux, Mac, Server, Cloud
Computer
Works best on a powerful computer or one with a graphics card (GPU)
Privacy
Runs on your own device: your data stays with you
API key
Not needed

Try it

Using Claude Code, Cursor or another AI coding assistant? Paste the prompt and it will walk you through the setup.New to this?

Before you install

  • Only download from the official GitHub page. Copies with similar names exist for many popular projects.
  • Plain-AyI is an independent guide and doesn't host or review the software it lists. Please check each project's official page before installing.
  • License (MIT): Very permissive: free to use, change and share, including for business.
  • More on staying safe

Where people are talking about it