# Pocket LLM for Android

An Android app that runs Gemma, Qwen and LLaMA language models entirely on the phone for offline, private chat, including voice input, image input with OCR and camera-based prompting.

Pocket LLM ships as a small base APK and lets the user download only the models they want to use, switching between them and deleting unused ones to save storage. It supports ONNX-based Qwen models and LiteRT-based Qwen 3 and Gemma 4 models, with streaming responses, persistent local chat history and Markdown-rendered replies.

Recent additions include voice input, image input with OCR and native image understanding, camera capture with retake and crop, a side panel for previous chats, editable model instructions with presets, and light and dark themes with adjustable font size. It runs fully offline after a model is downloaded, sends no telemetry, and is written in Kotlin using the LiteRT and ONNX Runtime backends.

Website: https://github.com/dineshsoudagar/local-llms-on-android