Google's Gemma 4 model goes fully open-source and unlocks powerful local AI - even on phones

Google DeepMind has announced the release of Gemma 4, its newest generation of large language models (LLMs), under the Apache 2.0 license. This marks a significant shift from previous Gemma versions, which used a restrictive 'Terms of Use' license, to a fully permissive open-source framework. The Apache 2.0 license grants developers and enterprises near-total freedom to use, modify, and redistribute the model for any purpose—personal, commercial, or enterprise—without royalties or Google's prior approval. This change unlocks powerful local AI deployment, allowing the model to run offline on a wide range of hardware, from cloud servers to smartphones and resource-constrained edge devices like Raspberry Pi.

The ability to run Gemma 4 locally addresses critical needs for data privacy, sovereignty, and operational resilience. Industries with strict confidentiality requirements, such as healthcare, can now leverage advanced AI capabilities without sending sensitive patient data to the cloud. Furthermore, devices in environments with intermittent or no network connectivity—common in IoT, manufacturing, and remote locations—can perform AI inference and decision-making in real-time, eliminating cloud latency and ongoing API costs. This move democratizes access to state-of-the-art AI, empowering a broader developer community to build and innovate on top of Google's research without the constraints of a closed ecosystem.