OfflineGPT is an experimental AI studio on your phone.
Chat with local models, connect your own API when you want, and explore OfflineGPT tools — including on-device Python (Power Mode). It should feel like a capable AI chat, but it’s built as a playground for people who like to experiment: try models, try tools, keep control.
WHAT YOU CAN DO
• Chat with AI models running on your device
• Download or attach compatible local models
• Connect your own API (OpenAI-compatible and similar)
• Optionally subscribe to Private Cloud for hosted models (Gemma 4 31B, Nemotron)
• Use experimental tools — including the Python sandbox in Power Mode
• Work without an account or required cloud login
OFFLINE BY DEFAULT
Local chat and local models stay on your device.
You choose when anything leaves the phone: your own API, or optional Private Cloud.
PRIVATE CLOUD (OPTIONAL)
Gemma 4 31B and Nemotron on Swiss infrastructure: 100% renewable power, server heat reused for heating.
Personal use with fair use — no token packs, no OfflineGPT account.
Monthly via Google Play — cancel anytime.
Local models and your own API keys work without it.
(Your own API provider may still bill their own usage.)
GOOD TO KNOW
• Local AI is slower than cloud AI — performance depends on your phone and model
• Best on modern devices (Android 12+, 6–8 GB RAM recommended; Android 10 / 4 GB minimum)
• Some tools and API features need network when you use them
• Experimental features can fail or produce imperfect results
• OfflineGPT is not a drop-in replacement for every cloud AI product
HOW TO START
1. Install the app
2. Download a model — connect your API — or use optional Private Cloud
3. Chat, explore tools, and build your own setup
Private offline AI playground — local models, chat, APIs & tools.