Introducing FunctionGemma

From the creator

Meet FunctionGemma, a specialized version of the Gemma 3 270M model, fine-tuned specifically for function calling and tool use and designed to be fine-tuned further. In this video, Kat Black (Product Manager, Google DeepMind) and Ravin Kumar (Research Scientist, Google DeepMind) introduce how this lightweight model enables developers to build fast, private, and cost-effective applications that translate natural language into executable API actions. Despite its small size (only 270 million parameters), FunctionGemma is designed to run efficiently on-device, making it perfect for mobile hardware, embedded systems, and edge computing. We also showcase two demos running on Google AI Edge: Mobile Actions that controls device features (like the flashlight and calendar) via voice, and an interactive farming game driven entirely by natural language commands. Resources: Announcing FunctionGemma, a specialized version of our Gemma 3 270M model that’s fine-tuned for function calling → https://goo.gle/44ViZ6n Train your own specialized agent with our Colab notebook → https://goo.gle/4p7SqCb Run your specialized Mobile Actions model in Google AI Edge Gallery → https://goo.gle/49jPJIK Subscribe to Google for Developers → https://goo.gle/developers Speakers: Kat Black, Ravin Kumar Products Mentioned: Gemma

Published 2025-12-19 · Watch on YouTube ↗
Availability and playback restrictions are controlled by the original channel.