Web and Node developers who want to run machine learning models client side without a backend.

A JavaScript port of Hugging Face Transformers that runs the same pretrained models in the browser or Node via ONNX Runtime.

Runs fully client side in browsers, browser extensions, or Node.js with no GPU server required.

Use when a product needs private, offline, or zero-latency inference such as in-browser embeddings, classification, or vision tasks.

It removes server costs and network round trips by shipping the model to the client.

Depends on ONNX Runtime Web and ships model weights compatible with the Hugging Face Hub.

Open source, Apache 2.0 licensed, maintained by Hugging Face; free to use with no credentials required.