Google releases Gemma 4, DeepMind's open-source models without usage restrictions
Google just dropped Gemma 4, its newest open-source AI model family from DeepMind, and this time there are no usage restrictions.
You can download it for free and run it locally (even offline), which is a big win for anyone who cares about privacy or needs to keep sensitive data on their own devices.
Think health care apps, or anywhere you want powerful AI without sending information to the cloud.
Gemma 4 server and phone variants
There are four versions: two (26B and 31B) built for high-powered servers, and two (E2B and E4B) made to run smoothly on phones or IoT gadgets.
All of them handle advanced reasoning, multiple languages, and visual tasks, while only E2B and E4B handle audio tasks.
So whether you're building something big or small, Gemma 4 lets you plug in smart features while keeping things efficient and more secure by not relying on cloud servers.