google/functiongemma-270m-it

TEXT GENERATIONConcurrency Cost:1Model Size:0.3BQuant:BF16Ctx Length:32kPublished:Oct 8, 2025License:gemmaArchitecture:Transformer1.0K Gated Cold

FunctionGemma is a lightweight, 270 million parameter open model developed by Google, built on the Gemma 3 architecture with a 32K token context length. It is specifically trained for function calling, enabling it to translate natural language into structured function calls. This model is designed for fine-tuning to create specialized function-calling agents, particularly for resource-constrained environments like mobile devices or local infrastructure.

Loading preview...

FunctionGemma: A Specialized Model for Function Calling

FunctionGemma, developed by Google, is a compact 270 million parameter model built on the Gemma 3 architecture, specifically engineered for function calling tasks. Unlike general-purpose dialogue models, FunctionGemma is intended as a foundation for creating highly specialized function-calling agents through fine-tuning. It leverages the same research and technology as the Gemini models, optimized for text-only function calling with a 32K token context length.

Key Capabilities

  • Function Calling Specialization: Translates natural language inputs into structured function calls, ideal for agentic workflows.
  • Lightweight and Efficient: Its small size (270M parameters) allows for deployment in resource-limited environments such as laptops, desktops, or mobile devices.
  • Fine-tuning Ready: Designed to achieve high performance after further fine-tuning on specific function-calling tasks, including multi-turn scenarios.
  • Versatile Deployment: Optimized for various hardware, supporting use cases like voice-controlled interactive games (Tiny Garden) and mobile OS actions (Mobile Actions).

When to Use This Model

  • Developing Custom Agents: Ideal for developers looking to build specialized agents that convert user commands into app-specific or system-level functions.
  • Resource-Constrained Applications: Suitable for on-device or edge deployments where computational resources are limited.
  • Offline Functionality: Enables the creation of private, offline agents for personal device tasks, as demonstrated by the Mobile Actions use case.
  • Enhancing Application Interactivity: Can power interactive experiences by handling game logic or triggering system tools based on user input.