# LocalMind

A single web page that runs several AI models entirely inside the browser using hand-written WebGPU kernels, offering chat, image generation, document question-answering and OCR.

LocalMind runs real AI models directly on the user's own device from a single web page, with the model downloading once, caching in the browser, and working offline afterward. It offers eleven local chat models ranging from a small model of about 140 megabytes up to a 27-billion-parameter model running entirely in the browser tab, plus Chrome's built-in Gemini Nano model, alongside on-device image generation, an experimental text-diffusion mode, document question-answering over uploaded files with multilingual search, and OCR that extracts selectable text and tables from images or scanned PDFs.

Everything runs on the user's GPU through WebGPU, and the only network request the page ever makes is an optional web search, sent only when the user presses a search button, using a search key the user supplies. Some models also accept images and audio for captioning, visual question-answering and voice input.

Website: https://localmind.naklitechie.com