Google announces Gemma 4 open AI models, switches to Apache 2.0 license - Ars Technica

This page announces Google's release of Gemma 4, the first major update to its open-weight AI models in over a year. The new models come in four sizes optimized for local usage: two larger variants (26B Mixture of Experts and 31B Dense) designed for GPU processing, and two smaller ones (2B and 4B) for mobile devices. Gemma 4 focuses on reduced latency and efficient parameter activation. In response to developer concerns about AI licensing restrictions, Google has switched from its custom license to Apache 2.0, offering greater freedom for developers using these open models.