On-device AI for Android. Run open-weight LLMs, Gemma 4, Stable Diffusion and Whisper dictation on your phone. Local features are free.
Published by CodeAlp
OfflineGPT is only available on Google Play. Updates come through the Play Store. APKs from other sites are not from us.
com.offlinegpt.app 64-bit ARM devices only
Local models run through C++ libraries called via JNI. They need no internet connection.
arm64-v8a (64-bit ARM) Which local models fit depends mostly on RAM.
Small 0.6B–1.5B GGUF models (Qwen3 0.6B Thinking, LFM2.5 1.2B, Llama 3.2 1B), Whisper Base dictation and ML Kit OCR. BYOK and hosted models work on any device.
3B–4B open-weight models (Qwen3 4B Thinking, Gemma 4 E2B Multimodal, SmolVLM2 Vision), Supertonic 3 TTS and local Stable Diffusion.
Gemma 4 E4B Multimodal (text, vision, audio) and Stable Diffusion on the Qualcomm Hexagon NPU.
Install the app, download a model, start a chat.
Get OfflineGPT from Google Play.
Pick a model in the built-in catalog or the Hugging Face browser. The smallest is about 640 MB. Use Wi-Fi.
Local chat, image analysis, image generation and dictation also work in airplane mode. Chats, files and recordings stay on the device unless you use a cloud feature.