FunctionGemma 270M

A lightweight 270M parameter model engineered for rapid function calling and structured tool use on edge hardware.

FunctionGemma 270M brings efficient tool use to the edge by optimizing the Gemma 2 architecture for structured JSON output. It fits constrained environments (mobile apps and IoT gateways) using a 540MB memory footprint. The model focuses on high-precision intent mapping: delivering sub-50ms response times for local-first workflows. It provides developers with reliable: low-latency function calling while eliminating the privacy risks of cloud-based LLMs.

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