SenseTime unveils full-scenario AI agent products for integrated cockpit and autonomous driving at Beijing Auto Show
The AMW Read
Incremental product launch within a well-known segment and company; updates SenseTime's automotive portfolio but does not fundamentally alter the competitive landscape.
SenseTime unveils full-scenario AI agent products for integrated cockpit and autonomous driving at Beijing Auto Show
At the 2026 Beijing Auto Show, SenseTime's autonomous driving subsidiary, SenseAuto, launched a suite of AI agent products targeting integrated cockpit and autonomous driving. The lineup includes the Sage Box (端侧 model + Sage OS + New Member stack), New Member 2.0 autonomous cabin AI agent, R-UniAD 2.0 generative driving system, and SenseAuto Go Robotaxi integrated cabin-driving solution, with a partnership with T3 Mobility to start Robotaxi trial operations this year.
This launch signals the convergence of foundation-model-driven AI agents with automotive verticalization, a pattern seen across Chinese AI players moving from cloud-based models to edge-deployed, vehicle-centric products. SenseTime's strategy exemplifies the "context-engineering moat" — using proprietary end-side models (the Sage model, already deployed on NVIDIA Orin X) to reduce cloud token costs to near-zero, claiming up to 30 yuan per vehicle per day in savings, while enabling real-time, always-on autonomous cabin behavior. This mirrors broader industry moves toward on-device AI to solve latency and cost bottlenecks in autonomous driving, a recurring pattern in the China auto AI market.
The R-UniAD 2.0 generative driving system, which integrates multimodal interaction, world model simulation, and reinforcement learning, aims to serve as a unified L2-to-L4 driving base, with partnerships including Dongfeng Motor. Together, these announcements position SenseTime as a serious contender in China's competitive smart cockpit and autonomous driving segments, potentially reshaping the economics of in-vehicle AI by driving down cloud dependency.
