Gasgoo Munich - On February 4, INFIFORCE officially launched its first "mind-native" compact bipedal robot: YUANZI.
Standing 95 centimeters tall and weighing 20 kilograms, the YUANZI features 25 degrees of freedom across its body. As INFIFORCE puts it, the design reflects a highly restrained choice.
The underlying logic is that INFIFORCE views this 95-centimeter stature as the "golden scale" for human-robot interaction. It looks up when you stand and meets your gaze when you sit. This eliminates any sense of intimidation, preventing family members like the elderly or children from instinctively backing away. Meanwhile, the 20-kilogram weight allows an adult to move it effortlessly, enabling it to navigate between sofas, coffee tables, and desks — truly "moving in" to your life.
In terms of core capabilities, YUANZI boasts three standout features: it is growable, ultra-free, and truly understands you.
By "growable," the company means YUANZI isn't a static device the moment you buy it; rather, it is an evolving entity capable of continuous development.
Image Source: INFIFORCE
To support this, INFIFORCE has equipped YUANZI with universal peripheral interfaces. Its hands and key components can be rapidly swapped or upgraded — including custom 3D-printed add-ons — for plug-and-play functionality. On the software side, official over-the-air (OTA) updates and an open platform ensure the robot's capabilities are continuously expanding.
The "ultra-free" aspect relies on the "Yuanli Marketplace," a forthcoming platform where users can create robot functions using zero-code shortcut commands. Modules created by users can be uploaded to the marketplace for sharing or even paid distribution.
"Truly understanding you" is powered by YUANZI's "Embodied Brain." This system integrates INFIFORCE's self-developed Hyper-VLA end-to-end multimodal model and a Causal World Model. It also taps into Alibaba Cloud's Qwen large language model, delivering 275 TOPS of edge computing power.
The Hyper-VLA architecture primarily provides the robot with multimodal perception and continuous learning capabilities. Meanwhile, the Causal World Model enables it to grasp the "why" of the physical world — understanding the mechanics of force, the properties of objects, and the consequences of actions.
