FunctionGemma and Hugging Face

Google’s high-efficiency Gemma models fine-tuned for structured function calling and hosted on the Hugging Face platform.

FunctionGemma utilizes Google’s 2B and 7B architectures to convert natural language into precise API calls. Hosted on Hugging Face, these models integrate with the Transformers library to bridge the gap between human intent and executable code. Developers use these weights to build schema-aware agents for specific tasks: database querying, IoT control, or real-time data retrieval (achieving high accuracy with low latency). This combination eliminates the need for massive proprietary models while maintaining strict control over structured JSON outputs.

https://huggingface.co/google/gemma-7b-it