Ant Group has announced a strategic partnership between its embodied AI subsidiary Robbyant and robotics developer Leju Robot. The collaboration focuses on developing AI models that can control physical robots in real-world environments and accelerating the commercialization of so-called embodied AI systems.
Embodied AI refers to artificial intelligence that operates in a physical body such as a robot, combining perception, reasoning and action. Instead of executing a single pre-programmed task, the goal is to develop systems capable of handling a broader range of activities across different environments and robot platforms.
According to the companies, the partnership combines Leju’s expertise in robot hardware, real-world use cases and operational data with Robbyant’s work on AI models for robotics. Together they plan to collect and curate large datasets of robot interactions and use these to train and refine AI systems that can perceive their surroundings, interpret instructions, make decisions and execute actions.
A key element of the collaboration is the development of high-quality training data from real robots. Leju previously supplied nearly 10,000 hours of multimodal robot interaction data to support the development of Robbyant’s LingBot-VLA model. This vision-language-action model is designed to serve as a general control system for robots, combining camera input, language understanding and physical actions.
The companies state that the model has already been adapted to robots from several manufacturers, demonstrating the ability to transfer capabilities across different robot types and physical configurations.
Robbyant earlier introduced a suite of embodied AI models intended to support robot perception, simulation and decision-making. These include LingBot-Depth, which focuses on spatial perception and depth sensing in complex environments, LingBot-World, a world model used to simulate and predict environmental dynamics, and LingBot-VA, a model designed to reason about actions through video prediction.
Through the new partnership, Robbyant and Leju aim to apply these technologies to industrial and commercial service scenarios. The companies plan to develop benchmark applications and reference deployments intended to demonstrate how embodied AI can be used in real operational settings.
The collaboration reflects a broader trend in robotics in which AI developers and robot manufacturers are increasingly working together to combine large-scale machine learning with physical robotic systems. The approach is widely seen as a necessary step toward more flexible and general-purpose robotic capabilities beyond narrowly defined automation tasks.
