Chat with a local model

The model downloads into your browser and runs on your own device. No server, no API key.
Downloaded once, then cached by the browser.
Runs with Transformers.js using WebGPU when available, otherwise CPU.