In Brief: Google releases Gemma 4, an open-source model for local AI applications. The model supports autonomous agents, multi-step planning, and audio-visual processing in over 140 languages – all without requiring specialized fine-tuning.
Google DeepMind has released Gemma 4 – a new family of state-of-the-art open-source models that enables artificial intelligence directly on local hardware. Available under the Apache 2.0 license, the model empowers developers to create intelligent agents and autonomous AI applications without specialized training.
Google’s AI Edge Team presents Gemma 4, a breakthrough in edge AI technology. The new model transcends the limitations of simple chatbots and enables the development of agents as well as fully autonomous AI applications that run directly on end devices.
Gemma 4 delivers impressive capabilities without additional fine-tuning: multi-step planning, autonomous behavior, offline code generation, and audio-visual processing. These features make the model a versatile tool for developers.
A particular strength is the model’s global orientation. With support for more than 140 languages, Gemma 4 opens access to a worldwide application spectrum. The integration of image processing significantly expands the possible use cases.
Published under the permissive Apache 2.0 license, the model is now available to developers. This marks an important step toward democratizing advanced AI technology on edge devices.