tool calling

A compact 270M parameter model built for high-speed intent parsing and local tool calling.

FunctionGemma 270M IT executes low-latency intent parsing directly on the edge. This 270-million parameter model fits on mobile hardware and local workstations: it handles structured tool calling without cloud dependencies. It translates user prompts into precise JSON schemas or function calls (ideal for offline AI agents). By offloading routing tasks from larger models, it reduces operational costs while keeping user data local.

https://huggingface.co/google/gemma-2-270m-it