# TinyAI

An Android app for running large language models entirely on-device, so a downloaded model can be used to chat fully offline with no API key, server, or data leaving the phone.

TinyAI is an Android app for running large language models entirely on-device. After a model is downloaded once, every response is generated on the phone's own CPU or GPU, so it keeps working in airplane mode and no conversation touches a server.

It runs on two local inference engines, Google's MediaPipe GenAI and LiteRT-LM, automatically selecting the right one for a given model and switching between CPU and GPU backends per model. It supports vision-capable models, streaming responses that can be stopped mid-generation, per-model system prompts, multi-session chat history stored in a local database, and Markdown and LaTeX rendering. It is written in Kotlin with Jetpack Compose and Apache-2.0 licensed.

Website: https://github.com/YashBhadange2006/TinyAI