Alibaba unveiled the open AI module RynnBrain for robotics

Alibaba launches a new AI model for robots

The Chinese technology giant introduced an artificial intelligence model specifically designed for robotic systems. It enables machines to perform tasks in the real world and competes with solutions from Google and Nvidia.

What is RynnBrain?
* Origin – developed by Alibaba’s DAMO Academy (Research Center of Alibaba).

* Purpose – enabling robots to interact with their environment.

* Key capabilities

* Understanding spatial relationships over time.

* Planning a sequence of actions to accomplish tasks.

* Creating maps of objects and predicting trajectories.

* Navigating tight spaces (e.g., in kitchens or on assembly lines).

The model is named RynnBrain. The description emphasizes that it can “build a map of the world” and “predict movement,” as well as operate in confined spaces.

Technical details
* Base – the large language model Qwen3‑VL.

* Versions – from a basic 2 billion-parameter version to an advanced MoE (Mixture‑of‑Experts) configuration.

* Availability – open source, freely distributed on Face and GitHub platforms.

Competition
* Google is developing Gemini Robotics‑ER 1.5.

* Nvidia offers Cosmos‑Reason2.

Alibaba aims to establish itself in the physical AI segment, where a battle between China and the United States is currently underway. Robotics is one of Beijing’s priorities, and humanoid robots are seen as key to dominance across various industries—from manufacturing to hospitality.

Open source vs closed source
Chinese companies often release AI models with open source code, while American players prefer keeping the most advanced technologies in closed systems. The strategy of open code in physical AI has traditionally been limited to academic projects (e.g., Stanford and UC Berkeley), but it is now actively applied by large corporations.