Huawei Cloud Releases and Open-Sources Ascend-Native Embodied World Model CloudRobo-WM1.2
Huawei Cloud officially released and open-sourced its Ascend-native embodied world model, CloudRobo-WM1.2, achieving top rankings in both the RoboTwin 2.0 and LiberoPro 2.0 benchmarks. The release also includes the model's training framework and Ascend computing best practices to accelerate commercial deployment. This launch strengthens China's domestic embodied AI and robotics ecosystem by providing a high-performance, localized computing and modeling foundation. With 80% of Chinese embodied AI companies already utilizing Ascend Cloud, this open-source release aims to transition embodied AI from laboratory research to large-scale industrial applications. CloudRobo-WM1.2 incorporates five core technological breakthroughs, including a PPA multi-modal world representation architecture and a native DualAct architecture for action generation. Additionally, the CloudRobo platform has gathered over 16,500 developers and supports various deployment scenarios spanning industry, power, and healthcare.
## BACKGROUND
Embodied AI refers to intelligent systems, such as robots, that can understand, interact with, and learn from the physical world through sensors and actuators. World models in this domain simulate physical dynamics and environmental changes, enabling robots to plan long-term tasks and execute actions more safely and accurately.