Run models on-device, use hosted AI, or connect your own providers.
Architecture
Choose the balance of privacy, speed, and capability that fits your workflow.
Runs completely offline on your device. Prompts, documents, and voice never leave your phone.
Instant access to frontier cloud models with zero training — including Gemma 4 31B on 100% renewable energy (Infomaniak) and top API models.
Connect directly to OpenAI, Anthropic, or Groq. OfflineGPT acts as the client without a middleman.
Reviews
“I'm running models up to 35B on my PC and this is the only app that gives me full access to these models on my server and local models on my phone itself. If you install Gemma 4 you'll get fast answers with STT and TTS as well as tool calls. I think its gotten better than PocketPal, Off Grid or LLM Hub.”
“It's one of the best offline AI's I've used. It runs smooth as butter on my device. It has so many options, features, and models to choose from. Upgrading to premium is optional, the app still works great without it.”
“Only downloaded it as an experiment; I have my own solutions for running AI. But as far as local AI running on your phone, I can't imagine needing more than this. On my Pixel 10 it runs pretty quick.”
“This app is awesome, it is the best LLM app that I've seen so far. It has everything you could need, literally everything — you have a bunch of tools and most of them are free. I've tried different apps but they're slower than this one.”
“This app is extremely well thought out and one of the best ways to run LLMs locally. The app works directly with the hardware and model you give it.”
“Dang this is the best offline AI model app I've ever tried. The fact that it selects the best AI model for you is amazing :D”
“This is an excellent app for running the various AI models. It is well designed and easy to use. Works well with the diffusion models. Thanks for making this available!”
“Gives the user a lot of control over their local AI, and in my books that makes this app really useful. And the UI is nice.”
FAQ
Yes. Once you download a local model, on-device chat (GGUF & Gemma 4 LiteRT), Stable Diffusion image generation, Whisper voice dictation, Supertonic 3 TTS, and ML Kit OCR run completely offline without internet.
On first launch, OfflineGPT automatically inspects your phone's available RAM and processor architecture to recommend quantized models that run smoothly without manual GGUF guesswork.
No account or registration is required. OfflineGPT Core is free: local models, voice, OCR, and local image generation have no feature paywall. Hosted cloud models are an optional add-on.
Chats and files are encrypted locally with SQLCipher (AES-256) on your phone. In BYOK mode, your API keys stay in the Android Keystore and connect directly to your provider. In Hosted mode, Gemma 4 31B runs on 100% renewable energy at Infomaniak with zero training.
Download
OfflineGPT on Google Play.
OfflineGPT · offlinegpt.tech · Developed by CodeAlp · com.offlinegpt.app
