Android
Google AI Edge Gallery
Best first stop for Android phones and tablets when the goal is a fast proof that Gemma 4 works locally.
Gemma 4 download
Download Gemma 4 through the path that matches your device.
This page is for the first click. Use the official app-store or runtime link first, then move into the device guide or runtime guide that fits the machine you actually own.
iPhone and iPad
Google AI Edge Gallery
Start with the App Store path first, then move into the iPhone and iPad guide if you need the realistic limits and fallback routes.
Mac
LM Studio
The easiest desktop download path for Apple silicon when you want a clean first local run before deeper runtime tuning.
Windows
LM Studio
Best default for most Windows laptops and RTX desktops when setup speed matters more than low-level control.
Runtime downloads
Pick the runtime that matches your intent, not just the loudest recommendation.
AI Edge Gallery
Phone-first Gemma 4 tests on supported mobile devices.
LM Studio
The shortest Mac and Windows path to a working local desktop chat.
Ollama
Local API workflows after you already know the model fits your machine.
llama.cpp
The highest-control path for troubleshooting, templates, and serious local tuning.
Before you download
The wrong first download is still the most common Gemma 4 mistake.
Phones should prove E2B or E4B before anything heavier.
16GB Macs should treat E4B as the safe start and 26B A4B as the stretch goal.
Desktop GPUs need context and memory budgeting, not just a bigger model file.
Try Online
Test Gemma 4 in the browser before you commit to a local setup.
Open try online →
Hardware Requirements
Map phones, 16GB Macs, and desktop GPUs to realistic starting models.
Open hardware requirements →
Compare
Compare model sizes and runtimes without jumping between disconnected pages.
Open compare →