Off Grid - Private AI Chat

Wednesday Solutions

About this app

Your phone is powerful enough to run AI. Why send your data to someone else's server?

Off Grid runs AI entirely on your device. No internet required. No accounts. No subscriptions. No data ever leaves your phone.

WHAT YOU CAN DO
- Chat with AI - Run Llama, Qwen, Gemma, Phi and other open models
- Generate images - Stable Diffusion running on your phone's NPU
- Vision AI - Point your camera at anything and ask questions
- Voice input - Whisper transcription, completely offline
- Attach documents - Analyze PDFs, code, text files privately

WHY OFF GRID
- 100% offline after downloading models
- Your conversations never leave your device
- No cloud costs, no API limits, no subscriptions
- Works on airplane mode
- Open source (MIT licensed)

PERFORMANCE
- 15-30 tokens/sec text generation on flagship devices
- 5-10 second image generation with NPU acceleration
- Supports GPU offloading for faster inference

PRIVACY FIRST
Use AI for sensitive tasks you'd never trust to the cloud - journal entries, health questions, work notes, personal photos. Your data stays yours.

Download a model, go offline, and never look back.

Ratings and reviews

Update: Reduced to 3 stars because it doesn't pick up all models on remote servers, nor their skills. Even very common ones. Awesome app. It has most of what I've been looking for that no other app has or at least has functioning correctly. It would be a lot better with the pro features, hopefully i can get those and contribute to the repo as well with some of my own ideas.

Thanks for the explanation! The local network is scanned when you first open the app, before you can configure anything, and regardless of the user's choice: whether your launch scan or not, the app is just constantly scanning local network (all the time apparently). I think it's not really optimal way. As a user, I would like to be able to control local network scanning.

When generating images with a Mediatek Dimensity 8500 Ultra (Mali-G720 MC8) and 12GB RAM it appears to do the one time GPU optimization every single prompt and doesn't save anything to GPU cache (0 when cleared). Is this because it only works properly with Snapdragon chips? Are you planning to support other chips than just the Snapdragon? It also doesn't show image previews, even when GPU acceleration is turned off.

What’s new

Faster on your hardware: Off Grid now uses your phone's GPU and NPU for quicker responses. Flagship Qualcomm devices auto-enable the NPU; others default to GPU.

New:
Aggressive Loading mode for max single-model performance.
"Load Anyway" when a model won't fit.
Fixes: voice transcripts never send empty, voice image requests route correctly, no more reasoning text leaking into replies, cleaner model cards, and fewer out-of-memory crashes.