Hall C · Booth C301

Beijing SynapX Dynamics Technology Co., Ltd.

Registered Chinese name北京章鱼灵动科技有限公司Beijing SynapX Dynamics Technology Co., Ltd.

Founded in early 2026, SynapX is a hand-brain integrated operation intelligent platform company focusing on Physical AI. It is committed to building AI native physical productivity…

"Data + model + dexterous hand" integrated design of rope-driven dexterous handUltimate senseless & full-modal large-scale embodied intelligent data acquisition product suiteSYNWorld Embodied Native Universal World Basic Model
Supply Chain & Components

Company Profile

Founded in early 2026, SynapX is a hand-brain integrated operation intelligent platform company focusing on Physical AI. It is committed to building AI native physical productivity with "universal world model + universal dexterous hand + full-modal data system". The core team brings together the world's top talents in AI large models, autonomous driving, embodied hardware and other fields, and has reached 100 people. The company completed several rounds of financing within 3 months of its establishment, with the cumulative amount reaching 1 billion yuan. The company will release three core products globally at WRC: tens of billions of native embodied world models; model and data collaborative design to benchmark Tesla's tendon-rope dexterous hands; and Ego, EgoBio, and DexUMI full-modal data acquisition solutions consisting of Ego visual headband, Bio myoelectric bracelet, and Dex tactile gloves. The company promotes the transformation of embodied intelligent productivity across the entire industry and builds the next generation of global physical AI infrastructure.

Key Products

03 ITEMS
"Data + model + dexterous hand" integrated design of rope-driven dexterous hand
01

"Data + model + dexterous hand" integrated design of rope-driven dexterous hand

The first high-degree-of-freedom, "data + model + dexterous hand" integrated design rope-driven all-wheel drive five-finger dexterity hand that can be mass-produced. The omnidirectional size is benchmarked against the adult male hand, with 21+ independently controllable degrees of freedom replicating the dexterity of the human hand; the 1900+ mechanical sensing unit array combined with joint back-drive technology approaches the human hand's perception capabilities; during the full cycle of 1000+ hours of trouble-free operation, the sub-millimeter level positioning accuracy approaches the stability of the human hand.

Ultimate senseless & full-modal large-scale embodied intelligent data acquisition product suite
02

Ultimate senseless & full-modal large-scale embodied intelligent data acquisition product suite

With Bio2Robot as the core concept, three major collection paths of DexUMI × EgoBio × Ego are constructed in layers to break through the problem of full-modal, large-scale, high-precision, and zero-interference collection of embodied data. The first dexterous hand isomorphic design bionic exoskeleton data mining glove achieves accurate capture of human operation data and efficiently transforms human operation experience into data assets that can be learned by robots. The hardware includes the core components of Ego visual headband, Bio myoelectric bracelet, and Dex tactile gloves, forming three major data acquisition solutions: Ego, EgoBio, and DexUMI.

SYNWorld Embodied Native Universal World Basic Model
03

SYNWorld Embodied Native Universal World Basic Model

The basic model of the embodied world with tens of billions of parameters is trained based on billions of high-quality samples. It has the ultimate universal generalization ability and can achieve extremely fast cross-ontology adaptation with only dozens of samples. The model adopts an innovative full-modal MoT architecture to achieve unified modeling of multi-dimensional core capabilities such as world model, inverse dynamics, value assessment, video-action joint prediction, and action-progress joint prediction. At the perceptual level, the model is natively compatible with pinhole cameras and fisheye cameras. Through 3D spatial geometric alignment, it seamlessly integrates multi-source perspectives such as head and end to achieve global spatial modeling of multi-view observation data.