Did this land?
Off Grid AI Mobile · iOS & Android

The whole studio,
in your pocket.

Off Grid AI Mobile runs real models on your phone - chat, vision, image, voice, and documents, all on-device. Point it at the Mac running Off Grid AI Desktop and it uses those bigger models over your own network, no relay. Pro adds a voice that talks back, personas, and draft actions you approve. Nothing leaves your devices.

Download on the App Store Get it on Google Play Star on GitHub

iPhone 12 or newer · Android 10+ · 4GB RAM · free to download

chat · vision · image · voice input · projects · tools · any GGUF · your Mac's models over your own network · and more

The AI on your phone logs every prompt to someone else's server. Off Grid runs the model in your phone's memory instead. Turn on airplane mode and it still answers. The whole thing stays on the device in your hand.

What you get for free

A complete offline AI suite on your phone. Not a chatbot - text, image, vision, voice, and documents, all running on your own hardware.

Chat
Text and vision, streaming, with a thinking mode. Qwen, Llama, Gemma, Phi, or any GGUF you bring. 15-30 tokens a second on a flagship phone.
Image generation
On-device Stable Diffusion with a live preview. NPU-accelerated on Snapdragon, Core ML on iPhone. 5-10s an image on a flagship.
Vision AI
Point your camera at anything and ask. Read a receipt, describe a scene, pull text off a document. On-device with SmolVLM, Qwen3-VL, or Gemma 3n.
Voice input
Hold to record and on-device Whisper turns your speech into text. No audio ever leaves your phone.
Projects
Drop in PDFs and docs. They are chunked and embedded on-device, then chat grounded in them with cited sources.
Tools
Built-in web search, calculator, date, and knowledge-base lookup, so a model that supports tool calling can act on live information.
Your Mac's models
Off Grid finds the Mac running the desktop app on your network, or any Ollama and LM Studio server, and runs their bigger models from your phone. Over your own LAN, never a relay.
Offline by default
Inference runs on the phone, so nothing round-trips to the cloud. Works in airplane mode, on the subway, on a plane, anywhere.
Download on the App Store Get it on Google Play Star on GitHub

Off Grid AI Pro: a voice, personas, and actions

The free app runs models on your phone. Pro is an optional, additive tier: it gives the assistant a voice that talks back, personas you shape, and the tools to draft real actions you approve. One license covers your phone and your Mac. All on-device.

Voice mode
Free gives you speech-to-text. Pro adds on-device text-to-speech with Kokoro, so it talks back and you run the whole thing hands-free. The voice runs in your phone's RAM.
Custom personas
Give each assistant its own system prompt, voice, and persistent memory, so it stays in character across conversations.
Draft, then approve
Connect Calendar, email, and MCP servers like Linear, Notion, and GitHub. It drafts the reply or files the ticket and waits. Nothing sends without your tap.
Sync, landing through July
Off Grid AI Sync is rolling out through July. When it lands, your phone and your Mac merge into one picture over your own network, never a relay. Your license includes it the day it ships.
Off Grid AI Pro is live: $49/year. The price climbs as we grow — never down — so today's tier is the lowest it will be. One license covers up to 5 devices, on mobile and desktop.
Get Pro — $49/year See everything Pro does

Why you can trust it

Privacy first by architecture. Your data never leaves your phone.

The model runs in the phone’s memory and answers on the phone’s own chips. There is no server to leak and nothing is logged. 100,000+ downloads across the apps, 2,500+ GitHub stars, a 500-strong community.

Star on GitHub

Questions

Is it really free? The full local studio is free and open source under MIT, and it keeps shipping. Pro is the optional add-on: voice output, personas, draft actions, and sync.

Does it work offline? Yes. Inference runs on the phone. Airplane mode, the subway, a plane, anywhere.

Which phones? iPhone 12 or newer on iOS 16+, and Android 10+ with 4GB of RAM or more.

Does it phone home? No cloud inference and nothing logged. Pro activates with a license key, not a cloud account. The Pro draft-action tools reach out only to the services you connect, and sync runs over your own network, never a relay.

What models can I run? Qwen, Gemma, Llama, Phi, and any GGUF small enough for your phone. Or connect to the Mac running Off Grid AI Desktop, or any Ollama or LM Studio server, and run their bigger models over your LAN.

What does Pro cost? $49/year today, climbing toward $99/year as we grow — the tier you join at is the rate you hold. Prefer to own it? $69 once today for lifetime. One license covers 5 devices, phone and laptop.

Download on the App Store Get it on Google Play Star on GitHub