Google has launched its latest artificial intelligence model, Gemma 4, marking a major shift towards more open and flexible AI tools for developers.
The new model, released on Thursday, is being offered under the Apache License 2.0. This means developers can freely use, modify and distribute it, including for commercial purposes, as long as proper credit is given.
Unlike many advanced AI systems, which are often restricted or tied to paid services, Gemma 4 is fully open source. It can run directly on personal devices, including Android phones and laptops, without needing a constant internet connection.
A different path from Gemini
The launch also highlights a clear difference between Gemma and Gemini AI, Google’s widely used AI platform.
Gemini powers several of Google’s core services such as Search, Gmail and Docs, and is typically accessed through cloud-based tools or subscriptions. Gemma, on the other hand, is designed to run locally and gives users more control over their data.
This local setup means conversations and files processed through Gemma 4 stay on the user’s device, offering a more private experience.
Built for reasoning and real-world tasks
According to Google, Gemma 4 is its most capable open model so far. It is designed to handle complex tasks that involve step-by-step thinking, planning and problem solving.
The model also brings improvements in maths and instruction-based tasks. It supports coding assistance, can process images and videos, and is able to convert speech into text.
Gemma 4 has been trained on more than 140 languages and supports large context windows, allowing it to handle long and detailed inputs more effectively.
Multiple versions for different needs
The model is available in four sizes, ranging from smaller versions designed for mobile devices to larger ones built for high performance systems.
The smaller variants are optimised to run on smartphones and lightweight hardware, while the larger versions can operate on powerful graphics processors such as those developed by Nvidia.
This range allows developers to choose a version that suits their hardware and project requirements.
Open models like Gemma 4 are gaining attention as they reduce reliance on expensive subscriptions and give developers more freedom to build their own tools.
Google said earlier versions of Gemma had already been downloaded hundreds of millions of times, with a large community creating thousands of variations.
With Gemma 4 now fully open source, the company appears to be encouraging even wider adoption, especially among developers looking for more control and flexibility in building AI-powered applications.
Read next: Apple introduces iPhone 17e
