Bring state-of-the-art agentic skills to the edge with Gemma 4- Google Developers Blog

Introduces Gemma 4, a family of open models for on-device AI development, enabling agentic use cases like multi-step planning, autonomous action, offline code generation, and audio-visual processing. Highlights Google AI Edge Gallery's 'Agent Skills' for on-device autonomous workflows, including knowledge base augmentation, content generation, and integration with other models. Details LiteRT-LM for deploying Gemma 4 across diverse hardware with minimal memory footprint, constrained decoding, and dynamic context. Provides performance metrics for Gemma 4 on Raspberry Pi 5 and Qualcomm Dragonwing IQ8. Announces a new Python package and CLI tool for Gemma experimentation and pipeline development.