China’s Alibaba released an open source AI model called RynnBrain, designed to power robots, enabling them to perform a variety of tasks.
The tech and e-commerce giant, through its DAMO Academy research initiative, made the model available on Feb. 10 on GitHub and Hugging Face, where it also explained its capabilities.
Describing RynnBrain as “an embodied foundation model grounded in physical reality,” the team behind the tech said in a statement that it went beyond “passive observation” to execute complex tasks using physics-aware reasoning.
An accompanying video demonstrated exactly what they meant by that. Headlined as RynnBrain’s Housework Diary, the three-minute-plus clip depicted a robot performing a number of domestic jobs on request.
Among these were arranging tableware around a sink, following specific instructions about what goes where, and identifying three oranges from a selection of fruit and placing them in a bowl. The robot also fetched a bottle of milk from a fridge, before putting a number of items away in an untidy front room.
While the tasks may seem relatively straightforward to human eyes, they require the RynnBrain-powered bot to master a number of different skills, including counting, spatial awareness and episodic memory, to identify objects, target areas and motion trajectory.
The result, according to Alibaba, could have significant consequences for how humans use and interact with bots in the future. “When robots have brains, they are no longer just tools, but think one step ahead of you,” stated a caption on the video.
RynnBrain was trained on Alibaba’s Qwen3-VL vision-language model, and, as with the vendor’s other models, its availability on GitHub and Hugging Face, in multiple versions, means developers can use it for free.
This has been a strategic tactic of Alibaba as it attempts to expand the use of its models.
But the tech giant is entering a competitive arena as the race for AI supremacy goes into overdrive. Among the big names it’s up against are Google and Nvidia, although Bloomberg reported that the Chinese giant is claiming to have successfully benchmarked its model against Google DeepMind’s Robotics-ER 1.5 and Nvidia’s Cosmos Reason2.
Elon Musk has also made no secret of Tesla’s ambitions to lead in AI and robotics, placing them at the heart of the company’s fourth master plan, published last year. But despite continual hype about its Optimus humanoid bot, tangible progress has been slow, with Musk recently admitting in Tesla’s 2025 Q4 earnings call that it is still “not in usage in our factories in a material way”.
