Qwen-3 became the first general-purpose AI model to be uploaded and operated in space after Chinese aerospace startup Adaspace Technology deployed it to an orbital computing center.

Adaspace executive vice-president Wang Yabo said the company completed multiple on-orbit inference tasks in November, according to Star Market Daily, SCMP cited on Tuesday.

Wang said the whole process from uploading queries from Earth, running inference in orbit, and transmitting results back to the ground took less than two minutes.

Star-Compute Project Scales Space AI Ambitions

Adaspace deployed Qwen-3 within its initial space computing center, part of the broader "Star-Compute Project." Wang said the project envisions a 2,800-satellite network designed to support physical AI as well as AI training and inference.

The initial computing center, launched in May last year, marked the world's first AI computing satellite constellation, consisting of 12 satellites, according to Securities Times.

Adaspace's second and third space computing centers are currently in production and scheduled for launch in 2026, Jiemian News reported.

Adaspace plans to deploy 2,400 inference satellites and 400 training units in low-Earth orbit, with full deployment expected by 2035.

China has designated aerospace a strategic priority through 2030 as it continues to compete closely with the U.S. in orbital launches.

BABA Price Action: Alibaba shares were up 1.68% at $174.25 during premarket trading on Tuesday, according to Benzinga Pro data.

Photo via Shutterstock