Google Releases ‘Gemma 4’ as Fully Open Source

Google has publicly released its new artificial intelligence model, ‘Gemma 4,’ under the fully open source Apache 2.0 license. Users can download the model to their own devices and operate it without internet access, helping to keep data secure and private. Gemma 4 is available in four sizes, ranging from 2 billion to 31 billion parameters, and supports more than 140 languages.

This time, Google has made significant progress by making Gemma 4 not just ‘open weight’ but fully ‘open source.’ Developers and users worldwide can download, modify, and use Gemma 4 for commercial purposes without paying any royalties. The key difference between Gemini and Gemma 4 is that Gemini is Google’s proprietary, paid technology, whereas Gemma 4 is an open model.

Key features of Gemma 4 include high capacity, multimodal processing, and availability in various sizes. It can process audio and video, perform speech recognition, and analyze charts. According to Google, Gemma 4 can easily run on billions of Android devices and general-purpose laptop GPUs. The Apache 2.0 license enables companies to maintain full control over their data and infrastructure while developing AI agents and coding assistants.