Every reply is generated entirely on your device — never by a company's server. Sign in free, download once, and it keeps working even offline.
Runs on the graphics chip already inside your device — the same one that powers your games. No cloud, no rented servers — the model never leaves your device to run.
Quick question? Fast answers in seconds. Complex problem? Thinking works through it step by step. Not sure? The app decides for you.
Every reply is generated locally, whether you're signed in or not. Sign in with Google only to sync your chats and settings across your own devices — never shared with anyone else.
Once a model is downloaded and cached, you can keep chatting without a connection.
No need to know anything about AI models or hardware specs — UnpluggedAI checks what your device can handle and loads the best version automatically.
Start for free. Upgrade for unlimited use of every model, billed once through Stripe.
Every reply shows which tier answered and why — the same chips you'll see inside the app.
What you see here is the real app — the same AI running in your browser right now is exactly what you get after signing up. Not a demo, not a simulation.
Write, summarize, answer questions, debug code, brainstorm, or talk through a problem — anything you'd use a general AI chat assistant for, generated entirely on your device.
It depends on the tier and your hardware — lighter models are a few hundred MB, the most capable ones a few GB. Download time depends on your connection, but it only happens once: after that, the model is cached on your device.
Any modern browser with WebGPU support — Chrome, Edge, and other Chromium-based browsers today, with Safari and Firefox support expanding. If your browser can't run it, the app tells you before you get stuck.
Yes. Signing in only identifies your account for your plan and settings — it doesn't change how replies are generated. Your conversations stay in your browser unless you're signed in, in which case they sync to your own private document that only you can access.
Lifetime means unlimited access for as long as UnpluggedAI is running — like any one-time software purchase, it's tied to the product staying live. We'll always aim to give existing users fair notice if that ever changes.
A private browser tab still sends your messages to that company's servers to generate a reply. UnpluggedAI generates every reply locally instead — no server is ever involved in answering you, whether you're signed in or not.
Running AI locally uses your GPU, so you'll notice more resource use while a reply is generating — similar to a demanding game. The app matches the model to your hardware to keep this manageable, and usage stops between messages.
To run real models here you need Safari on iOS 17.4+ / macOS Sonoma+, or a recent Chrome/Edge on Android or desktop. Update your browser or open this file on another device.
Pick a tier, or let the router choose for you.Models run for real on your GPU via WebLLM. The first reply on each tier takes longer because it downloads and loads the model; after that it's cached.