Free. On your device.

A small AI model runs in your browser. No AI-provider bill, server inference or API key. Replies can be wrong; this model has no web access or action tools.

Experience

Modes change guidance and controls, not model power or account privileges. Shared agents and enterprise services remain planned.

Local model

Qwen 2.5 · 0.5B · about 290 MB of model files, plus runtime downloads. Allow roughly 1.1 GB GPU memory; device needs can vary.

Download hosts: Hugging Face and GitHub. They receive normal download connection information, not your chat prompts. Model files may remain cached on this device. Downloads can use mobile data.

I agree to download the model to this device.

Checking this browser…

Chat storage & model files

Messages stay in this open page by default. Optional saved chat is browser storage, not encrypted or account-synced; anyone using this browser may read it. Only recent turns are used as model context.

Remember this chat on this device

Model & runtime licenses

STARTER / LOCAL CHAT

A useful first question.

Load the model, then ask a short question. Your prompts and replies are processed on this device.