Shangtang (00020) shareholder meeting reveals: Enterprise-level Token monthly usage has increased 7 times rapidly.

date
18:00 25/06/2026
avatar
GMT Eight
Since its official launch in May 2026, the enterprise token usage of Megvii's "DailyNew" large model has grown rapidly, increasing by 7 times in a single month.
SenseTime Technology (00020) publicly disclosed the latest operational data of its enterprise-level Token Plan for the first time at the shareholders' meeting: since its official launch in May 2026, the usage of the "RiRiXin" large model enterprise-level Token has rapidly increased by 7 times within a single month. SenseTime attributes this growth to the collaborative efforts of three core capabilities: its SenseNova 6.7 Flash-Lite lightweight model, which reduces Token consumption costs by 60% to achieve extreme low costs; the open office skill support for "one-click deployment" of production-level workflows, significantly reducing the threshold for enterprise access; and the millisecond-level real-time response capability that fully adapts to high-frequency production office scenarios. The scale growth of the SenseTime Token Plan benefits from the deep cost reduction of the software and hardware integration, achieving a healthy business loop. In terms of product ecosystem data, it was disclosed that the total number of users of the desktop intelligent agent "Raccoon" has reached 20 million, serving over tens of thousands of corporate clients, with a weekly active user (WAU) of 2.1 million; the AI content creation platform Seko serves over 800,000 individual creators (a fourfold increase year-on-year) and over 1,300 corporate clients - the continuous growth of these two product lines confirms the commercial reality of the RiRiXin large model in the broader office and content production scenarios. Leveraging its own cluster of 40,000+ Peta FLOPS of computational power, SenseTime achieved positive cash flow in the second half of 2025. The meeting also disclosed that SenseTime's SenseNova-U1 Pro, a native full-modal intelligent agent base with the industry's first "understand-generate-act" unified core, benchmarking GPT-Image 2, will be launched in the second half of 2026, focusing on visual design as the main track, and is expected to start invite testing in July. This is also seen as the next ignition point for accelerating the commercialization of SenseTime's large models.