Get Pollen for Android

Pollen runs Gemma 4 E4B on a pocket device via Google AI Edge LiteRT-LM, with native multimodal audio. Two build flavors — pick the one that matches what you want to test.

Demo flavor

Walk through the app, no model needed

~56 MB APK · no model · runs on Android 12+

Deterministic contract evaluator, no 3 GB model on disk. The launcher icon carries a "D" watermark and the app shows a permanent "DEMO APP — NO LLM" banner. Dependency-injected stub clients so it works offline-first, with no Jetson or ESP32 powered on.

Download pollen-demo.apk

Direct download from zigiella.com. File lands in your phone's Downloads folder.

Device flavor

Run real Gemma 4 E4B on your phone

~102 MB APK + 3.6 GB model side-loaded · Android 12+ / API 31+

Real LiteRT-LM pipeline. Side-load gemma-4-E4B-it.litertlm (3.6 GB) once and grant microphone permission. Then the app runs Gemma 4 E4B fully offline on your phone. Requires 12 GB RAM recommended, 6 GB free storage.

Download pollen-device.apk

Direct download from zigiella.com. After download, side-load the Gemma 4 E4B model — instructions below.

Install on Android (one-time setup)

  1. Open Files or Downloads on your phone and tap the downloaded APK.
  2. Android asks: "For your security, your phone isn't allowed to install unknown apps from this source." Tap Settings.
  3. Enable "Allow from this source" for Chrome (or whichever browser you used).
  4. Go back and tap the APK again. Tap Install.
  5. Open the app from the launcher.

Device flavor: also side-load the model

After installing pollen-device.apk, push the model file to the phone via ADB:

# From a computer with adb installed:
adb push gemma-4-E4B-it.litertlm /data/local/tmp/gemma-4-E4B-it.litertlm

On first launch of the app, tap "Skip (Model already on device)" and confirm the path /data/local/tmp/gemma-4-E4B-it.litertlm. Grant microphone permission to enable voice input.

Full step-by-step in the Pollen build guide.