What makes local AI possible
Browser-based AI relies on WebGPU, a modern standard that lets web pages use your graphics card for heavy computation. With WebGPU plus enough graphics memory, a browser can load and run language models on-device — fully private, with no API costs and no data leaving your computer.
The main constraint is memory. Tiny models (around 1 billion parameters) run on most modern laptops; mid-size models need a few gigabytes of graphics memory; the largest models want a dedicated GPU with 6 GB or more. This tool estimates where you land.
Why on-device AI matters
On-device AI is private by default — your prompts and documents never leave your machine — and free to run once the model is downloaded. That's why local-first apps are increasingly building AI features that work offline and cost nothing per use. Checking your hardware is the first step to using them.