Local AI.
Zero API keys.
Llama, Qwen and Gemma running on your GPU, in your browser. Chat, query your PDFs, compare models and run OCR — prompts and documents never leave your device.
100% client-side · WebGPU + WebAssembly · no API keys · no trackers
drop a PDF and ask — local RAG with page citations
Llama / Qwen / Gemma on your GPU
compare two local models side by side
on-device image OCR & captioning (Florence-2)
How is this possible?
Modern browsers expose the GPU through WebGPU. ai0 loads 4-bit-quantized LLMs with WebLLM (MLC) and vision/embedding models through Transformers.js, compiling and running them on your own hardware. The only network traffic is the one-time model download from the model CDNs — everything else is local. Open DevTools → Network and watch the silence.