A humanoid robot that can see the future, at least a few seconds of it, just debuted in China. Unitree Robotics unveiled its UnifoLM-X2-1.0 model on September 7, claiming it to be the world’s first fully autonomous humanoid robot powered by a real-time world model.
The distinction matters. Most humanoid robots today either follow pre-scripted behaviors or require a human operator pulling the strings remotely. Unitree says its new system does neither. Instead, the robot uses a world model to predict how physical interactions will unfold and then plans its movements accordingly, all on the fly.
What a world model actually does
For a humanoid robot, this means it can anticipate how objects will move, how forces will interact, and how its own body will respond to different actions. Unitree specifically highlighted “high-dynamic actions” in its demonstration, suggesting the robot can handle fast, complex physical tasks that would normally require human reflexes or teleoperation.
The company released a video demonstration alongside the announcement. Unitree described the robot as capable of combat-level autonomy, a provocative framing that positions it well beyond the warehouse-pick-and-place tasks most robotics companies are chasing.
This isn’t Unitree’s first move in the world model space. The company open-sourced its UnifoLM-WMA-0 framework back in September 2025, a precursor designed to improve decision-making in robotic systems. The X2-1.0 model appears to be the production-grade evolution of that research, tackling the harder problems of real-time planning and dynamic execution.
From startup to public company in a decade
Unitree was founded in 2016 by Wang Xingxing in Hangzhou. The company started with quadruped robots, essentially robotic dogs, before moving into the humanoid space.
As of July 2026, Unitree had shipped more than 18,000 humanoid units. The company turned profitable, then went public on the Shanghai STAR Market in August 2026. That IPO raised approximately $905 million and was oversubscribed.
Wang has consistently emphasized vertical integration as a competitive advantage, designing and manufacturing core components in-house to keep costs down. The results show in Unitree’s pricing. Its G1 humanoid models sell for roughly $16,000 to $18,000, a fraction of what competitors like Boston Dynamics or Figure AI charge for their platforms.
The competitive landscape heats up
What separates Unitree’s latest move is the claim around full autonomy. Most competitors still rely heavily on teleoperation for complex tasks, meaning a human operator controls the robot remotely while machine learning handles simpler, repetitive functions.
Wang Xingxing has pointed to applications spanning education, entertainment, and industrial automation as the primary targets.
The $905 million from the IPO gives Unitree significant capital to scale manufacturing and continue investing in AI research. Wang has signaled that substantial portions of that funding will go toward AI model development.