Alibaba releases its open source RynnBrain robotics foundation model to give machines perception, reasoning and action skills, aiming to help researchers and startups build smarter robots without proprietary lock-in.
Alibaba has released RynnBrain, an open source embodied foundation model designed to serve as a cognitive “brain” for robots, shifting foundational robotics intelligence into an open development ecosystem.
Built on Alibaba’s Qwen3-VL, the model enables machines to perceive, reason and act in physical environments, moving beyond purely digital AI tasks. The company said it supports “performing physically aware reasoning and executing complex real-world tasks”, along with spatial mapping and environmental understanding to help downstream Vision-Language-Action (VLA) systems execute more sophisticated actions.
By open sourcing the stack, Alibaba allows researchers, robotics startups and hardware makers to build and customise robot intelligence without proprietary lock-in. The move positions open development as a catalyst for faster progress in humanoid robots, factory automation and smart devices.
Industry observers view the spatial reasoning capabilities as a technical step forward. Charlie Zheng, chief economist at Samoyed Cloud Technology Group Holdings, said, “The distinctive spatial reasoning capability marks a leap for Chinese developers in embodied intelligence foundational models.”
The release also addresses a longstanding bottleneck in robotics: dependence on pre-programmed routines. General-purpose reasoning and action could accelerate commercial deployment across real-world settings.
Elsewhere in China’s AI ecosystem, DeepSeek expanded its large language model context window from 128,000 tokens to more than one million, enabling longer memory and improved reasoning, though it was not described as open source.
Separately, South Korea unveiled a 1 trillion-won ($692.6 million) five-year initiative to develop 10 on-device AI semiconductors for applications including self-driving cars, smart homes and humanoid robots, aiming to reduce reliance on foreign chipmakers.
