Wang Xingxing: Unitree G1 cumulative production approximately 10,000 to 20,000 units, physical AI expected to see breakthroughs
Wang Xingxing, founder of Unitree Technology, revealed that since its release in May 2024, the Unitree G1 humanoid robot has cumulatively rolled off the production line at approximately 10,000 to 20,000 units, with a starting price of 99,000 yuan, a height of about 127 cm, and a weight of about 35 kg. He believes that physical AI is expected to make key progress in the next one to three years, and that the "ChatGPT moment" of embodied intelligence may arrive as early as 2027, but the core challenge remains the alignment between AI models and the real physical world.
Body · AI translation
On October 22, Wang Xingxing, founder of Unitree Robotics, revealed at the 2026 World Physical AI Conference and Unitree Partner Ecosystem Conference that since its release in May 2024, the Unitree G1 humanoid robot has cumulatively produced approximately 10,000 to 20,000 units off the line, making it one of the best-selling humanoid robots in the world.
As a representative product under Unitree, the G1 stands about 127 centimeters tall, weighs about 35 kilograms, and starts at 99,000 yuan, mainly targeting fields such as scientific research and education, artificial intelligence development, and robot application exploration.
Speaking about the development prospects of physical AI, Wang Xingxing said that the world currently still places insufficient emphasis on artificial intelligence, and that including himself, people have underestimated the development potential of AI. Physical AI has now shown signs of technological breakthrough, and key progress is expected within the next one to three years, with the "ChatGPT moment" of embodied intelligence possibly arriving as early as 2027.
Coinciding with the tenth anniversary of Unitree Robotics' founding, Wang Xingxing showed on site a robot test video based on the GPT-6 large model. In the video, the robot can complete operations such as drawing, path planning, and navigation without additional training for specific tasks, demonstrating the potential of large models in robot task understanding and execution.
However, Wang Xingxing pointed out that the core challenge for embodied intelligence to move toward large-scale application remains the alignment problem between AI models and the real physical world. Although large models already possess strong planning, navigation, and error-correction capabilities, when facing unfamiliar objects, fine manipulation, and complex environments, robots are still limited by factors such as sensor errors and execution precision, and the actual task success rate needs to be improved.
In terms of technology research and development, Unitree is conducting preliminary research on the Physical AI Robot Self-Evolution V1.0 large model, exploring ways to let AI autonomously participate in code writing, data collection, testing, and model iteration, and continuously optimize system capabilities through a small amount of human feedback. Wang Xingxing expects that as the model's autonomous evolution capability continues to strengthen, technological progress in physical AI is expected to further accelerate.
Submissions/leaks: tougao@youxituoluo.com
Manuscripts/business cooperation: Lin Nan (WeChat 19250561593)
Join the industry exchange group: Lin Nan (WeChat 19250561593)
Metaverse Digital Industry Service Platform
Download the "Tuoluo Technology" APP to get cutting-edge in-depth metaverse information