Skip to content
Get packs

brain.js

A JavaScript library for building small neural networks that run in browsers and Node.js.

GitHub Download 1.4.10
Stars
14.9k
Forks
1.1k
Licence
MIT
Language
TypeScript
Last update
2.1 years ago
Type
Library

No commits in over a year — check for forks or alternatives before adopting.

What it is

brain.js is a JavaScript library for creating and training neural networks, which are programs that learn patterns from example data. It runs in both web browsers and Node.js, and its description says it is GPU accelerated. It is a continuation of the older harthur/brain project, which is no longer maintained.

  • Train a small feed-forward network to classify inputs such as colors into categories, as in the README's color contrast demo
  • Train a recurrent network on sequence data and use forecast to predict upcoming values
  • Train a network offline and export it with toJSON or toFunction to use on a website
  • Stream training data using the Train Stream feature listed in the README

Great for

  • JavaScript developers who want to try neural networks without learning a new language
  • Developers adding simple pattern recognition, such as classifying inputs, to a web page
  • Students learning how neural networks train and make predictions
  • Teams that want to train a model offline and then run it in the browser with toJSON or toFunction

Not ideal for

  • Large-scale deep learning or research that needs Python tooling and large datasets; the README does not compare performance with other frameworks
  • Projects that need the latest model types such as large language models; the README describes only classic network types like feed-forward, RNN, LSTM and GRU
  • People who need a guaranteed install on every machine, since the README notes GPU support depends on a native module that may need building from source
Step-by-step guide: how to use brain.jsInstall, first run, examples, tips and FAQs

How to get it

  1. 1

    Install with a package manager

    You'll need: Node.js (version not specified in the README)

    Install from npm

    npm install brain.js
  2. 2

    Run it

    • Install the package with npm install brain.js
    • Create a network with const net = new brain.NeuralNetwork(config)
    • Call net.train() with an array of input and output objects
    • Call net.run() with a new input to get a prediction
  3. 3

    Get the most out of it

    • Use the activation option with values such as sigmoid, relu, leaky-relu or tanh to change how neurons respond
    • Set hiddenLayers to change the network's size
    • Train offline and use toJSON or toFunction to deploy a trained network in the browser
    • Use the Train Stream and Cross Validation features described in the README for larger workflows

Downloads

Source code (.zip)master branch · straight from GitHubRelease 1.4.10Release notes and sourceOfficial website & docsbrain.js.org

Strengths

  • Simple API: create a network, call train with an array of input and output pairs, then call run
  • Works in both browsers and Node.js, with a CDN script option for quick use
  • Has a documented set of network types and training options, plus examples and a demo site

Trade-offs

  • Installing GPU support depends on the native headless-gl module, which may need Python, compilers or OpenGL drivers when prebuilt binaries are unavailable
  • Training is described in the README as computationally expensive, so it is best done offline or in a worker
  • Per the README it is a continuation of an unmaintained project, and the latest release shown is from a repository last pushed in September 2024

How it compares

ProjectStarsTypeLevelUpdatedBest for
brain.js (this)14.9kLibrarybeginner2.1 years agoJavaScript developers who want to try neural networks without learning a new language
mocha22.9kLibraryintermediate5 days agoJavaScript and Node.js developers who want automated tests for their code
turndown11.5kLibrarybeginner19 days agoDevelopers who need to convert HTML content, such as web pages or rich text editor output, into Markdown
next.js143kFrameworkintermediate4 days agoDevelopers already comfortable with React who want a full framework around it
debug11.5kLibrarybeginner22 days agoNode.js developers who want named, toggleable log output instead of plain console.log calls
page-agent29.6kLibraryintermediate6 days agoDevelopers adding an AI copilot to their own web app

Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.