OpenmindAGI/functiongemma-finetuned-g1-multilingual

TEXT GENERATIONConcurrency Cost:1Model Size:0.3BQuant:BF16Ctx Length:32kPublished:Apr 13, 2026License:apache-2.0Architecture:Transformer Open Weights Cold

OpenmindAGI/functiongemma-finetuned-g1-multilingual is a 0.3 billion parameter FunctionGemma model, fine-tuned by OpenmindAGI, designed to convert natural language into structured robot action and emotion function calls. This model supports six languages (English, Chinese, Japanese, French, German, Spanish) with 98% accuracy and achieves approximately 59ms inference speed on NVIDIA Jetson AGX Thor. Its primary strength lies in enabling multilingual human-robot interaction by translating diverse natural language inputs into predefined robot commands and emotional responses.

Loading preview...

Model Overview

OpenmindAGI/functiongemma-finetuned-g1-multilingual is a specialized 0.3 billion parameter model, fine-tuned from Google's FunctionGemma 270M, to translate natural language commands into structured robot actions and emotional expressions. This model is particularly optimized for real-time human-robot interaction, demonstrating high accuracy and fast inference speeds on edge devices.

Key Capabilities

  • Multilingual Support: Processes inputs in 6 languages: English, Chinese, Japanese, French, German, and Spanish.
  • Function Calling: Converts natural language into specific robot actions (e.g., shake_hand, face_wave, do_payment) and emotional states (e.g., happy, sad, confused).
  • High Accuracy & Speed: Achieves 98% accuracy with an inference time of approximately 59ms on NVIDIA Jetson AGX Thor, utilizing constrained decoding for significant speedup over standard autoregressive generation.
  • Efficient Training: Fine-tuned using LoRA (rank 8, alpha 16) on a dataset of ~6,000 examples, including diverse multilingual phrases generated via Claude API to ensure natural expression.

Ideal Use Cases

This model is well-suited for applications requiring:

  • Robotics: Enabling intuitive, multilingual voice or text control for robots.
  • Interactive Systems: Developing systems where natural language commands need to be translated into predefined actions or states.
  • Edge AI Deployments: Its small size and fast inference make it suitable for deployment on resource-constrained hardware like the NVIDIA Jetson AGX Thor.

Finetunes
1 models