Alibaba's DAMO Academy launches RynnBrain, an open-source AI model that enables robots to perform real-world tasks
Alibaba has released ' RynnBrain ,' an open-source platform model for robots and autonomous devices. RynnBrain is a model that gives robots the ability to 'see, understand, and act' in the physical world, and is expected to accelerate the development of physical AI.
RynnBrain
China's Alibaba launches AI model to power robots
https://www.cnbc.com/2026/02/10/alibaba-ai-model-robotics-rynnbrain-china.html
DAMO Academy, a cutting-edge research institute within Alibaba Group, a major Chinese IT company, has released a video showing a robot identifying fruit and placing it in a basket. While the task may seem simple at first glance, the robot must understand each fruit, pick it up with the correct amount of force, and place the correct number of fruit in the desired location. This requires active task execution skills that take into account the laws of physics.
RynnBrain is a foundational model rooted in physical reality that empowers robots with 'physical intelligence' through comprehensive egocentric cognition, which allows them to understand their environment from a self-centered perspective, accurate spatiotemporal reasoning, and real-world behavior planning. By leveraging not only spatial understanding from video but also spatiotemporal reasoning based on memory, including past observations, RynnBrain can accurately identify movement trajectories and target regions.
On the official RynnBrain page, you can see several demonstrations of its capabilities. Below is the task of 'putting a plate back to its original position,' which requires object recognition, positional understanding, and spatial memory.
In addition, in a task where the robot sorts food into categories, it accurately recognizes the objects and lifts them with the appropriate force.
Below are the benchmark results presented by Alibaba. Alibaba reports that RynnBrain's performance showed cutting-edge results compared to
RynnBrain is available in various versions with 2 to 30 billion parameters, and the 30 billion parameter model has shown competitive results in some robotics-related benchmarks, including comparisons with large-scale language models such as GPT-5.2 and Gemini 3 Pro.
As China competes with the United States for dominance in AI development, it is placing emphasis on robotics, including humanoid robots, as a key pillar of its strategic plan to reshape various industries with physical AI. RynnBrain marks Alibaba's first entry into the robotics market, and is seen as a way for the company to take the lead in robotics AI by encouraging developers and researchers around the world to improve the technology through an open-source strategy.
RynnBrain is an open source model and can be downloaded from GitHub or Hugging Face.
GitHub - alibaba-damo-academy/RynnBrain: RynnBrain: Open Embodied Foundation Models
https://github.com/alibaba-damo-academy/RynnBrain
RynnBrain - a Alibaba-DAMO-Academy Collection
https://huggingface.co/collections/Alibaba-DAMO-Academy/rynnbrain
Related Posts:
in AI, Posted by log1e_dh