Alibaba eyes physical world with its first suite of AI models for robots

Alibaba Group Holding has launched the Qwen Robot Suite, its first set of AI models tailored for robotic applications, marking a shift toward embodied intelligence. The suite, unveiled in Shanghai, aims to move AI beyond chatbots by enabling machines to perceive, reason, and interact with physical environments. Developed by Alibaba’s AI research division, Tongyi Lab, the models are already undergoing pilot testing with select Alibaba Cloud enterprise clients. The Qwen Robot Suite is structured into three core components. Qwen-RobotNav is a vision-language navigation model that helps robots understand and navigate physical spaces. It works alongside Qwen-RobotWorld, a video-based world model that allows robots to predict and simulate how environments will change before acting. Physical execution is handled by Qwen-RobotManip, a generalist vision-language-action model built on the Qwen3.5-4B architecture. The suite represents Alibaba’s latest effort to expand AI capabilities beyond digital interactions into real-world applications. By integrating perception, reasoning, and action, the models aim to enhance robotic autonomy in dynamic settings. The pilot phase with Alibaba Cloud clients will test the models’ performance in practical scenarios, paving the way for broader adoption. Alibaba’s move underscores the growing competition in embodied AI, where companies seek to develop robots capable of complex physical tasks. The Qwen Robot Suite’s layered approach—combining navigation, simulation, and manipulation—highlights the company’s focus on bridging AI and robotics. This initiative aligns with the broader industry trend of embedding AI into physical systems for enhanced functionality.