Alibaba Debuts Qwen-Robot Series of Embodied Intelligence Models
As embodied intelligence technology moves from labs into real-world applications, the ability for robots to accurately understand instructions and autonomously perform tasks in complex environments has become a key industry focus. On June 16, Alibaba officially launched the Qwen-Robot series of embodied intelligence large models, giving various robots a “general-purpose foundation” that can comprehend natural language, perceive three-dimensional spaces, and understand physical laws.
The Qwen-Robot series includes three core models that can either operate independently or work together, forming the first complete embodied intelligence matrix within the Qwen family.

First is Qwen-RobotManip, responsible for core manipulation. To solve the problem of traditional models suffering significant performance drops when switching robot platforms, this model adopts a unified action representation and has been pre-trained on over 38,000 hours of open-source data. In third-party authoritative evaluations, its different versions not only secured the top two positions in task success rates but also demonstrated breakthrough capabilities in complex, high-difficulty tasks—from simply turning on a tap to performing dual-arm French fry flipping.
Next is Qwen-RobotNav, which gives robots the ability to “navigate” and “run errands.” This model unifies five navigation functions—task instruction understanding, target search, and autonomous driving—within the same framework. Its innovative “task-adaptive observation mechanism” enables robots to completely move away from rigid memorization strategies, allowing them to “walk, look, and plan” flexibly while efficiently completing object-finding tasks in complex unknown spaces.
Finally, the Qwen-RobotWorld model enhances the depth of a robot’s “thinking.” This is a physical world model that, like an athlete rehearsing movements, can simulate the next moment’s physical state and actions. It not only effectively overcomes the bottleneck of insufficient training data but also allows robots to perform trajectory simulations before executing actions, ensuring absolute accuracy in physical operations.
Related article
Former Spotify Exec Launches Vinyl Bar Startup in Shibuya to Redefine Music Apps
As generative AI continues to reshape the creative landscape, a new wave of startups is empowering users to produce music through AI prompts, mimicking the style of any artist. A Vinyl Bar in Shibuya, a startup with an indie band-inspired name, is de
600B Sparse MoE Activates 27B, Boosting Open-Source Flagship Cost-Effectiveness
Step 5 Preview, the latest flagship foundation model, is now officially available. Its purpose is clear: a primary model engineered for real-world Agentic tasks. It addresses the longstanding "Pareto problem" in AI, where capability, efficiency, and
JD Cloud Unveils MiniMax M3 Large Model, Marking a Leap in Inference Efficiency
JD Cloud’s JoyBuilder platform has integrated the newly launched MiniMax M3 model, making it available to users immediately upon release.This update focuses on enhanced inference performance. The platform leverages its proprietary inference framework
Related Special Topic Recommendations
Comments (0)
0/500
As embodied intelligence technology moves from labs into real-world applications, the ability for robots to accurately understand instructions and autonomously perform tasks in complex environments has become a key industry focus. On June 16, Alibaba officially launched the Qwen-Robot series of embodied intelligence large models, giving various robots a “general-purpose foundation” that can comprehend natural language, perceive three-dimensional spaces, and understand physical laws.
The Qwen-Robot series includes three core models that can either operate independently or work together, forming the first complete embodied intelligence matrix within the Qwen family.

First is Qwen-RobotManip, responsible for core manipulation. To solve the problem of traditional models suffering significant performance drops when switching robot platforms, this model adopts a unified action representation and has been pre-trained on over 38,000 hours of open-source data. In third-party authoritative evaluations, its different versions not only secured the top two positions in task success rates but also demonstrated breakthrough capabilities in complex, high-difficulty tasks—from simply turning on a tap to performing dual-arm French fry flipping.
Next is Qwen-RobotNav, which gives robots the ability to “navigate” and “run errands.” This model unifies five navigation functions—task instruction understanding, target search, and autonomous driving—within the same framework. Its innovative “task-adaptive observation mechanism” enables robots to completely move away from rigid memorization strategies, allowing them to “walk, look, and plan” flexibly while efficiently completing object-finding tasks in complex unknown spaces.
Finally, the Qwen-RobotWorld model enhances the depth of a robot’s “thinking.” This is a physical world model that, like an athlete rehearsing movements, can simulate the next moment’s physical state and actions. It not only effectively overcomes the bottleneck of insufficient training data but also allows robots to perform trajectory simulations before executing actions, ensuring absolute accuracy in physical operations.
Former Spotify Exec Launches Vinyl Bar Startup in Shibuya to Redefine Music Apps
As generative AI continues to reshape the creative landscape, a new wave of startups is empowering users to produce music through AI prompts, mimicking the style of any artist. A Vinyl Bar in Shibuya, a startup with an indie band-inspired name, is de
600B Sparse MoE Activates 27B, Boosting Open-Source Flagship Cost-Effectiveness
Step 5 Preview, the latest flagship foundation model, is now officially available. Its purpose is clear: a primary model engineered for real-world Agentic tasks. It addresses the longstanding "Pareto problem" in AI, where capability, efficiency, and
JD Cloud Unveils MiniMax M3 Large Model, Marking a Leap in Inference Efficiency
JD Cloud’s JoyBuilder platform has integrated the newly launched MiniMax M3 model, making it available to users immediately upon release.This update focuses on enhanced inference performance. The platform leverages its proprietary inference framework





Home






