Unifollm-OminiA-0.3, a general intelligent model for humanoid robots, was recently released by ZOOX Technology. This model supports multimodal interaction and understanding, and can coordinate tasks across multiple scenarios such as home and elderly care, achieving a complete closed-loop from environmental perception, task understanding to autonomous execution.

According to the demonstration video released by ZOOX Technology, the humanoid robot G1 equipped with this model can perform various practical tasks, including moving a floor pillow to a sofa, identifying the color and quantity of items, precisely picking up a specified medicine box, organizing clothes, placing a dish in a dishwasher, and adjusting the height of a hospital bed.

QQ20260721-174936.jpg

In terms of interaction capabilities, Unifollm-OminiA-0.3 supports visual recognition and language understanding, enabling the robot to plan complex tasks based on user instructions. For example, the robot can identify three boxes of medicines of different colors and retrieve the medicine box at the specified location. Meanwhile, during device control, when users issue commands such as "stop," the robot can immediately halt its actions.

This model represents an important advancement for ZOOX Technology in integrating AGI (Artificial General Intelligence) and physical AI technologies on the humanoid robot G1 platform, further enhancing the robot's adaptability and task execution capabilities in real environments.

As the embodied intelligence industry develops rapidly, competition in the robotics field is shifting from purely hardware performance to a comprehensive capability of "model + perception + action." The release of Unifollm-OminiA-0.3 indicates that domestic humanoid robots are accelerating toward multi-scenario autonomous applications.