LM Playground
An Android app for running large language models locally on-device, with document Q&A, vision input, tool use and background model downloads.
LM Playground is an Android application for running large language models locally on-device, powered by llama.cpp with GGUF-format models from Hugging Face. Models are downloaded, loaded with one tap, and used entirely offline with no cloud calls or API keys.
It supports document question-answering over PDF, Word, EPUB, HTML, Markdown and plain-text files using on-device semantic search, vision input for photos through vision-capable models, and tools that let a model search the web, fetch a page or run JavaScript mid-reply, each disabled by default and enabled per model. Reasoning models display their thinking steps in a styled section, and system prompts can be saved and reused as personas.
It uses a custom download engine with resumable background downloads, supports loading custom GGUF model files, and is optimized for arm64 devices with KleidiAI kernels and OpenMP. It is written in Kotlin and licensed under MIT.