Multiple Models.
Eight optimised models across iOS and Android, from compact 1.5B assistants to record-breaking large models no consumer device has ever run before, including a 24B model on iPhone Pro, a 30B Mixture-of-Experts on Android, and a 70B model on flagship Android hardware that sets a new benchmark for on-device AI. SmolLM3 ships built-in. Download the rest on demand.


Works Anywhere.
Tube, plane, countryside, basement. LiberaGPT runs entirely on your iPhone's processor. Ask questions and get answers instantly without waiting for a network connection.
Your Data Stays Yours.
Delete your history instantly. Nothing leaves your phone. No cloud sync, no telemetry, no training data harvesting. When we say private, we mean it.

What's New.
Four new features in this release. All on-device, all private, all iOS.
Everything runs natively on your iOS or Android device.
on iPhone, LiberaGPT uses quantised GGUF models with A19 Neural Engine and Metal GPU acceleration, giving you fast, intelligent AI that actually respects your privacy.
Completely Offline
Works on the Tube, on a plane, in the countryside. No Wi-Fi required. Ever. Your conversations stay between you and your phone.
Neural Engine Optimised
A19 Neural Engine and GPU acceleration via Metal on iPhone 17 Pro tier devices. Quantised models (Q4_K) balance quality and speed. Fast responses with minimal battery impact.
Zero Data Shared
Nothing is sent to our servers. Nothing is sent to Apple. Your conversations are saved only on your device, and you can clear them any time.
Download What You Want
Choose which AI models to install. Download lightweight models for quick answers, or specialised models for deeper reasoning tasks.
No Recurring Charges
Pay once. Use forever. No subscriptions, no token limits, no surprise bills. For serious work, use cloud AI. For everything else, use LiberaGPT.
Nature Friendly
Data centres consume massive power. Your phone uses watts, not kilowatts. Get simple answers locally instead of burning energy in a server farm.

