4 things devs should know about Gemma 4

Google has released Gemma 4, its latest open model family, in four sizes: 2B, 4B, 26B (mixture of experts), and 31B. Key highlights include state-of-the-art performance for their sizes, purpose-built support for agentic workflows with native tool use and large context windows (up to 256K tokens), multimodal edge processing with audio and vision support on smaller models, and release under a permissive Apache 2.0 license. The models were developed by Google DeepMind with the same security protocols as proprietary models.