Another Better, Faster, Cheaper from a Big AI Wizard Type

October 16, 2025

This essay is the work of a dumb dinobaby. No smart software required.

Cheap seems to be the hot button for some smart software people. I spotted a news item in the Russian computer feed I get called in English “Former OpenAI Engineer Andrey Karpaty Launched the Nanochat Neural Network Generator. You Can Make Your ChatGPT in a Few Hours.” The project is on GitHub at https://github.com/karpathy/nanochat.

The GitHub blurb says:

This repo is a full-stack implementation of an LLM like ChatGPT in a single, clean, minimal, hackable, dependency-lite codebase. Nanochat is designed to run on a single 8XH100 node via scripts like speedrun.sh, that run the entire pipeline start to end. This includes tokenization, pretraining, finetuning, evaluation, inference, and web serving over a simple UI so that you can talk to your own LLM just like ChatGPT. Nanochat will become the capstone project of the course LLM101n being developed by Eureka Labs.

The open source bundle includes:

A report service
A Rust-coded tokenizer
A FineWeb dataset and tools to evaluate CORE and other metrics for your LLM
Some training gizmos like SmolTalk, tests, and tool usage information
A supervised fine tuning component
Training Group Relative Policy Optimization and the GSM8K (a reinforcement learning technique), a benchmark dataset consisting of grade school math word problems
An output engine.

Is it free? Yes. Do you have to pay? Yep. About US$100 is needed? Launch speedrun.sh, and you will have to be hooked into a cloud server or a lot of hardware in your basement to do the training. A low-ball estimate for using a cloud system is about US$100, give or take some zeros. (Think of good old Amazon AWS and its fascinating billing methods.) To train such a model, you will need a server with eight Nvidia H100 video cards. This will take about 4 hours and about $100 when renting equipment in the cloud. The need for the computing resources becomes evident when you enter the command speedrun.sh.

Net net: As the big dogs burn box cars filled with cash, Nanochat is another player in the cheap LLM game.

Stephen E Arnold, October 16, 2025

Written by Stephen E. Arnold · Filed Under AI, News, Open source, Smart Technology

Comments

Got something to say?

Search the site
Subscribe to Beyond Search
Feature archive
News archive

Stephen E. Arnold monitors search, content processing, text mining and related topics from his high-tech nerve center in rural Kentucky. He tries to winnow the goose feathers from the giblets. He works with colleagues worldwide to make this Web log useful to those who want to go "beyond search". Contact him at sa [at] arnoldit.com. His Web site with additional information about search is arnoldit.com.