ALIBABA'S QWEN-3 BECOMES ONE OF THE WORLD'S FIRST AI MODELS TO OPERATE IN ORBIT

Artificial intelligence has long depended on massive ground-based data centers, but that assumption is beginning to shift.

- 

In a development that underscores how quickly space and computing are converging, Alibaba Cloud has successfully run its Qwen-3 artificial intelligence model in orbit, completing real inference tasks aboard an Al-equipped satellite constellation. The test marks one of the first known instances of a general-purpose large language model operating in space rather than relying on Earthbound infrastructure.

The achievement places China at the forefront of a new category of computing, where data is processed directly in orbit instead of being transmitted back to ground stations. For industries ranging from climate monitoring to communications and national security, the implications are significant.

FROM DATA CENTERS TO LOW-EARTH ORBIT

The successful in-orbit operation of Qwen-3 represents a departure from the traditional model of Al deployment. Instead of satellites acting merely as data collectors, forwarding raw information back to Earth for processing, Qwen-3 performed inference tasks directly in space.

According to reports from Chinese media, the process was remarkably fast. From the moment a query was uploaded from Earth to the return of processed results, the entire workflow took less than two minutes. That speed highlights one of the core advantages of orbital Al: reduced latency and a dramatic reduction in bandwidth demands between satellites and ground stations. As satellite networks grow denser and sensors generate ever-larger volumes of data, sending everything back to Earth becomes increasingly inefficient. Running Al models in orbit allows only the most relevant insights to be transmitted, easing pressure on communications links.

THE ADASPACE COMPUTING CONSTELLATION