Scheduling Efficiency Up 80%: SenseTime Platform Backs First Agents in Shanghai Quantum City

SenseTime's large-scale computing platform, known as its large-device infrastructure, has supported the first batch of AI agents deployed in Shanghai's Quantum City initiative, with scheduling efficiency improving by roughly 80%, according to a report from QbitAI.

The report frames the shift as moving from understanding space to getting work done. That phrase captures a change in what the agents do: rather than only interpreting spatial information, they now take on actual tasks within the city's systems.

Quantum City points to scenarios where urban space and digital systems are tightly coupled. Agents operating there have to understand spatial information, business rules and live status at the same time, then make scheduling and execution decisions inside those constraints.

The concrete number in the report is the 80% improvement in scheduling efficiency. That figure speaks to task orchestration and resource scheduling: when multiple agents must cooperate inside one urban space and system, the efficiency of the scheduling layer largely determines whether the whole approach can run at scale.

SenseTime's large-device platform plays the role of the compute and platform foundation. Keeping agents running continuously in a city-scale scenario requires stable model inference, compute scheduling and engineering support, which is precisely what this infrastructure product is built to provide.

For SenseTime, the deployment is an outward demonstration of its platform in an urban setting. Whether agents can run reliably in a real city environment and produce measurable efficiency gains is a direct test of how mature an AI infrastructure stack really is.

The candidate material does not disclose how many agents make up the first batch, which business scenarios they cover, or how the 80% scheduling figure was measured, so those details will have to come from official follow-up.

Three things to watch: whether these agents move from a first batch to routine operation; whether SenseTime's platform can replicate the model in more cities and industries; and what quantifiable business value agents that get work done will deliver in practice.