TensorFlow.js

TensorFlow.js is Google’s open-source JavaScript library for training and deploying machine learning models in the browser and in Node.js. It brings the TensorFlow ecosystem to the web with hardware acceleration through WebGL, WebGPU and WebAssembly, so models can run entirely on the user’s device. First released in 2018, the project has accumulated about 19,100 stars on GitHub and is still actively maintained in 2026.

Overview

The library is organized as a family of packages: a low-level Core API for numerical computation and neural networks, a high-level Layers API modeled on Keras, a Data API for loading and preparing inputs, a Converter for importing existing TensorFlow SavedModels and Keras models, and Vis for in-browser visualization. Multiple backends let the same code run on the CPU, WebGL, WASM, WebGPU, native TensorFlow under Node.js, or React Native. Because inference happens client-side, TensorFlow.js enables private, low-latency ML features, such as image classification, pose estimation or in-browser retraining on sensor data, without sending data to a server.

Key Features

• Train and run ML models directly in the browser or in Node.js with a familiar TensorFlow API
• Hardware acceleration via WebGL, WebGPU and WebAssembly backends
• High-level Keras-style Layers API plus a flexible low-level Core API
• Converter to import pre-trained TensorFlow and Keras models and run them client-side
• Retrain existing models in the browser using local sensor or user data
• Rich ecosystem of pre-trained models, examples, tutorials and a project gallery

💳 Pricing

TensorFlow.js is completely free and open source under the Apache-2.0 license. Models run on your users’ devices or your own servers, with no licensing or inference fees.

Who Is It For

TensorFlow.js is for web and JavaScript developers who want to add machine learning to websites and apps, especially where privacy, offline use or zero server cost matters.