Galbot and Unitree Robotics Announce Joint Development of Humanoid Robot Galbot Star Kid

On October 22, at the 2026 World Physical AI Conference and Unitree Partner Ecosystem Conference, Galbot and Unitree Robotics announced that the world's first native intelligent agent robot, Galbot ET1 (Galbot Xingzai), was jointly developed by the two companies. In August this year, Galbot ET1 made its debut at the World Robot Conference. The announcement of joint development at the Unitree Partner Ecosystem Conference further reflects the deep collaboration between the two parties around product definition, hardware refinement, and embodied intelligence capability development. At the conference, Yao Tengzhou, co-founder of Galbot, was invited to deliver a keynote speech, focusing on how Galbot ET1, through real-time motion learning, multimodal interaction, and task planning, enables humanoid robots to evolve from executing preset actions to learning new skills, understanding new tasks, and acting autonomously. Yao Tengzhou, co-founder of Galbot, was invited to deliver a keynote speech. As physical AI moves toward reality, what kind of intelligence do robots need? Robots are becoming an important vehicle for artificial intelligence to move from the digital world into the physical world. Wang Xingxing, founder of Unitree Robotics, said at the conference that as foundation model capabilities continue to improve, physical AI may be approaching an important breakthrough moment. NVIDIA founder Jensen Huang has repeatedly emphasized that physical AI will become the next wave of artificial intelligence development. Tesla CEO Elon Musk has also expressed positive expectations for the industrial scale of humanoid robots. These judgments point to the same question: when AI begins to enter the real world, what capabilities do robots need? Wang He, founder and CTO of Galbot, proposed that the development of humanoid robots should not stop at action execution, but should further possess the ability to learn autonomously and move autonomously, be able to learn new skills from observation, and understand tasks and take action in real environments. This is also an important direction for Galbot's exploration of embodied intelligence. Galbot ET1 is precisely the product practice of this concept. Through real-time motion learning, multimodal interaction, and task planning, Galbot hopes to enable robots not only to complete preset actions, but also to observe the environment, learn new skills, understand tasks, and act autonomously according to actual conditions, gradually expanding the boundaries of capabilities in the real physical world. Complex movements learned on the spot, Galbot ET1 is the only one of its kind globally. What makes Galbot ET1 different from other humanoid robots? An intuitive answer is: its movements are more agile, and it is better able to directly learn new movements by watching. Behind this is the AstraBrain-WBC general cerebellum foundation model independently developed by Galbot; and what supports the continuous improvement of its capabilities is the million-hour-level human motion data accumulated by Galbot—including both high-precision motion capture data and motion data extracted and retargeted from human videos on the internet. With the help of the Galbot AstraData data foundation, Galbot continues to accumulate and utilize high-quality motion data to support the training of the general cerebellum model. On this basis, Galbot ET1 has demonstrated a globally unique learning capability: without the need to pre-produce new motion capture data, and without the need to provide pre-recorded motion videos, it can observe human movements in real time through vision, identify and extract motion trajectories, and learn and reproduce new movements on the spot. Furthermore, Galbot ET1 is gradually evolving intelligent-agent-level learning capabilities. In real scenarios, it can combine visual perception and language interaction to understand the environment and task requirements, plan action plans, and invoke capabilities such as movement and interaction to proactively communicate with people and attract customers. This means that robots are beginning to combine perception, motion, interaction, and task planning, no longer limited to completing a single action, but able to understand tasks and organize behavior according to the on-site environment, exploring more autonomous ways of interaction and task execution. Multiplying advantages: the industrial synergy between Galbot and Unitree Robotics. The birth of Galbot ET1 stems from the two parties' shared judgment on the development direction of humanoid robots: to break through the boundaries of robot capabilities, it is necessary not only to continuously refine key technologies, but also for all parties in the industrial chain to leverage their strengths and carry out deep collaboration around complete products. At the conference, Chen Li, co-founder of Unitree Robotics, said that as robot application scenarios continue to differentiate, the demand for complete machines in different scenarios is also changing, and it is necessary to jointly develop with partners and form industrial synergy. Yao Tengzhou, co-founder of Galbot, also said that the two parties share the same goal and each has its own strengths, and should work together to promote industry progress. The Galbot ET1 released today is precisely the next-generation humanoid robot redefined after the two companies each pushed what they do best to the extreme. Around Galbot ET1, Galbot deeply defines product design, continuously refines multimodal interaction and full-body motion control, and integrates embodied intelligence technology into complete product design and development. Unitree, meanwhile, leverages its accumulation in joint modules and hardware mass production to support hardware performance and reliability. The two parties began collaborating from the product definition stage, jointly refining overall machine performance, intelligent interaction, and scenario adaptation capabilities. The two parties "multiply" their respective strongest capabilities together, rather than simply adding them. This concept of collaborative innovation also echoes the direction proposed in the recent "Opinions on Promoting the Deep Integration of Technological Innovation and Industrial Innovation" issued by the General Office of the CPC Central Committee and the General Office of the State Council, namely building an enterprise-led, multi-entity collaborative innovation system integrating industry, academia, and research, and promoting the seamless connection between the innovation chain and the industrial chain. For the embodied intelligence industry, from core technology breakthroughs to product development and scenario application, it is necessary for upstream and downstream players in the industrial chain to further strengthen collaboration, so that technological accumulation in different fields can be transformed into complete product capabilities. Galbot will also continue to deepen cooperation with Unitree Robotics and more ecosystem partners to promote collaborative technological innovation and the development of the embodied intelligence industry. Exploring more possibilities for embodied intelligence. Not long ago, Galbot ET1 officially announced its pricing, launching a 79,000 yuan basic version, a 139,000 yuan professional version, and a 179,000 yuan flagship version, respectively targeting different needs such as motion display and interactive experience, active learning and dynamic planning, as well as scientific research and development and enterprise customization. After opening pre-orders, Galbot ET1 received more than 200 pre-orders within 24 hours. At present, Galbot ET1 has entered multiple cities across the country, leaving footprints in Hong Kong, Shanghai, Beijing, Guizhou, and other places, exploring application scenarios such as commercial display and interactive experience. The launch of the pricing system and the continuous expansion of real scenarios mean that humanoid robots are no longer just products displaying technology on exhibition stands, but have also begun to explore more specific application value for different user needs. I See. I Learn. I Act. From observation to learning, and from learning to action, this is the direction of intelligent evolution defined by Galbot ET1, and also Galbot's expectation for the future of embodied intelligence: to let AI step out of the digital world and enter the real physical environment, to learn and grow in continuous interaction with the physical world, to continuously expand the boundaries of capabilities, and to explore a future of infinite possibilities.