Google Releases Gemma 4 Open AI Models and Switches License to Apache 2.0

Google has launched Gemma 4, the latest generation of its open-weight AI model family, and simultaneously switched the licensing terms from the more restrictive Gemma-specific license to Apache 2.0, significantly expanding how developers and organizations can use, modify, and distribute the models, Ars Technica reported. The Apache 2.0 license change is potentially as significant as the model improvements themselves. The previous Gemma license placed restrictions on commercial use and redistribution that made it more difficult for enterprises and startups to build products on top of the models. Apache 2.0 removes most of those restrictions, allowing Gemma 4 to be used freely in commercial applications, fine-tuned, and redistributed as part of other products. Gemma 4 is built on the same research and architecture advances that power Google's flagship Gemini models, providing an open-weight alternative that developers can run locally, on private infrastructure, or in any cloud environment without API access restrictions. This is particularly valuable for organizations with data privacy requirements that preclude sending sensitive data to external AI APIs. The model family includes variants optimized for different hardware configurations, from single-GPU deployments to larger multi-GPU setups, allowing developers to choose the capability-to-compute tradeoff that suits their applications. The release comes as competition among open-weight AI models has intensified. Meta's Llama family, Alibaba's Qwen, and Mistral's models have all released competitive open-weight options in the past year, and Google's license change signals it intends to compete more aggressively for developer adoption in the open ecosystem rather than ceding that space to competitors.