Ant Group’s open-source push aims to move robots from lab demos to real-world work
Flashy robots can dance and backflip on cue, but Ant Group is betting open-source AI can take them beyond scripted moves
Chinese fintech giant Ant Group has open-sourced its first artificial intelligence models for robotics, as it steps up efforts to build machine intelligence capable of handling complex real-world tasks.
The move signals a deeper push by the Hangzhou-based Ant into embodied intelligence – AI systems designed to perceive, reason and act in physical environments rather than purely digital settings. The company is the fintech affiliate of Alibaba Group Holding, owner of the South China Morning Post.
Ant’s robotics arm, Ant Lingbo Technology, also known as Robbyant, said its latest releases include LingBot-VLA, a vision-language-action (VLA) model aimed at supporting what it described as a “universal brain” for robots, allowing more scalable and practical deployment across industries.
“For embodied intelligence to achieve large-scale adoption, we need highly capable and cost-effective foundation models that run reliably on real hardware,” said Robbyant CEO Zhu Xing. “Our goal is to accelerate the integration of AI into the physical world so it can deliver practical value sooner.”
China already leads the world in the deployment of industrial and humanoid robots.
However, many high-profile humanoids – including those developed by Unitree Robotics – that can dance and perform flips still rely on preprogrammed routines, limiting their autonomy and ability to generalise across tasks.
Chinese AI experts and investors have pointed to these limitations as a crucial obstacle to turning the new wave of robots into economically productive machines, making the development of a more powerful robotic “brain” a pressing priority for the global robotics industry.