AI Tools

How to Install and Run Gemma 4 on Android: Complete Getting-Started Guide

Install Google AI Edge Gallery on Android 12+, download Gemma 4, verify offline inference, and measure the device with the built-in benchmark.

If you want to run Gemma 4 offline on your phone, this guide walks you through the full process from setup to practical usage.

Step 1: Get the App

Google AI Edge Gallery is now available on Google Play and requires Android 12 or newer. Use the official listing first:

Only use the official GitHub Releases page when Google Play is unavailable. Avoid third-party APK mirrors and disable “Allow from this source” after installation.

On your Android device, go to:

Settings -> Apps -> Special app access -> Install unknown apps

Then:

  1. Find your browser (for example, Chrome or Firefox) and enable “Allow from this source.”
  2. Open the Google AI Edge Gallery GitHub Releases page in your mobile browser.
  1. Download the latest .apk package.
  2. After the download completes, open the file from notifications or your file manager and follow the prompts.

Installation and model download time depend on the connection, version, and model size; there is no reliable fixed two-minute estimate.

Step 2: Open the App and Grant Permissions

When you first open AI Edge Gallery, it will request storage permission to save model files. It’s best to allow this; otherwise, the app cannot download or load models.

You will typically see these sections on the home screen:

  • Ask Image: Vision tasks (describe images, answer questions about photos)
  • AI Chat: Standard text chat
  • Summarize: Paste text and generate summaries
  • Smart Reply: Generate reply suggestions

For most users, AI Chat is the primary entry point.

Step 3: Download a Gemma 4 Model

  1. Enter AI Chat.
  2. Tap Get Models when prompted.
  3. Choose a Gemma 4 model from the list (model size is shown).
  4. Pick based on your device capability; if your phone has 8GB RAM, start with Gemma 4 4B.
  5. Tap Download and let it run in the background.

Note: Larger models take longer to download. You can download multiple models and switch between them later. Downloaded models stay on your device, so you do not need to re-download them.

Step 4: Start Chatting

After the model download is finished:

  1. Tap the model name to load it (the first load usually takes 10 to 30 seconds depending on model size and device performance).
  2. Enter your prompt in the chat box and send it.
  3. The model generates responses locally, and your data does not leave the phone.

The first reply is often slower due to model warm-up. Later messages in the same session are usually faster.

Step 5: Try Vision Features (Gemma 4 Multimodal)

If you downloaded a Gemma 4 multimodal variant:

  1. Go back to the main menu and open Ask Image.
  2. Select an image or take a photo.
  3. Ask a question (for example, “What’s in this image?” or “Is there any text I should pay attention to?”).
  4. Wait for the model to analyze the image locally and return a result.

The model inference can run on-device; verify any external Skill separately because it may still access the network.

Benchmark and Offline Acceptance

Record the phone, SoC, Android and app versions, exact model and quantization, available memory, time to first token, sustained token/s, and any crash or thermal throttling. Run the same benchmark twice: the first pass includes loading and warm-up; the second better reflects sustained performance.

After the model downloads, enable airplane mode and run a fixed prompt. If it still generates, the core inference path is local. External Agent Skills may still require a network connection.

If a model will not load, free memory and try a smaller variant. If generation crashes or the phone becomes very hot, reduce context and stop the test. For APK failures, verify Android 12+, the official source, and whether a differently signed older build is installed.