Illustration of what gallery does: Android and iOS app for running open LLMs offline on your phone

google-ai-edge /

gallery

Android and iOS app for running open LLMs offline on your phone

  • 25.1k stars
  • 252 stars since 3 Oct
  • Kotlin
  • Updated today
  • Beginner friendly
  • Models & local AI
Open on GitHub (opens in a new tab)

AI Edge Gallery is a showcase app from Google. You install it on a phone, download an open model, and run it locally. The README says inference happens on the device, so no internet is needed once the model is downloaded.

The app is built on Google AI Edge and LiteRT, a lightweight runtime for running models efficiently on mobile hardware. Models are found and downloaded through Hugging Face, and you can also load your own. Each feature is a tile: chat, image questions, audio transcription, a prompt workspace, and a few experiments built on small FunctionGemma finetunes.

Unlike a cloud chatbot, your prompts and images stay on the device. It is also a testing bench: it has a benchmark feature that shows how each model runs on your phone. The README calls it an experimental beta, so expect rough edges.

What it does

  • Chat with Thinking Mode

    Multi-turn chat with a toggle that shows the model's step-by-step reasoning. It works with supported models, starting with the Gemma 4 family.

  • Agent Skills

    Adds tools such as Wikipedia lookups, interactive maps and visual summary cards to the model. You can load skills from a URL or browse community ones on GitHub Discussions.

  • Ask Image and Audio Scribe

    Ask questions about photos from the camera or gallery. Transcribe and translate voice recordings with on-device models.

  • Prompt Lab

    A workspace for single-turn prompts, with controls for settings like temperature and top-k.

  • Model management and benchmarks

    Download models from a list or load your own, then run benchmarks to see how each performs on your hardware.

What you could build

  • Try Gemma 4 on your phone without sending data to a server.
  • Compare how different open models perform on your own device with the benchmark tool.
  • Test prompts and sampling settings such as temperature and top-k in Prompt Lab.
  • Transcribe and translate voice recordings offline.
  • Build and load your own Agent Skills to give a local model extra tools.

Get started

Check that your device runs Android 12 or iOS 17 or later, then install the app from Google Play or the App Store. Without Google Play, download the APK from the latest GitHub release. A macOS download is also offered, and the project wiki has detailed guides. To build it yourself, see DEVELOPMENT.md.

Commands are copied from the project's README. Read it on GitHub for the full setup.

More in Models & local AI

See all
transformers Shared model definitions for text, vision, audio and multimodal AI models 168k open-webui Self-hosted chat interface for Ollama and OpenAI-compatible models, with RAG and plugins 156.5k2.7k odysseus Self-hosted workspace for AI chat, agents, research, documents, email and calendar 94.2k5.8k minimind Train a tiny 64M-parameter language model from scratch on one consumer GPU 64.3k1.2k Kronos Pre-trained Transformer models that forecast candlestick (K-line) price data 43.2k3.4k airllm Run very large open language models on a small GPU by loading one layer at a time 40.4k5.1k Vane Self-hosted AI answer engine that searches the web and cites its sources 37.7k681 timesfm Pretrained Google model that forecasts time series without training on your data 34.7k668 Unlimited-OCR Open OCR model that turns images and multi-page PDFs into structured text in one pass 28.2k1.6k lingbot-map Feed-forward model that builds 3D scenes from video frames as they stream in 18.2k989

    ↑↓ to move, Enter to open, Esc to close