[SenseTime Launches New Generation Lightweight Multimodal Intelligent Model with 60% Reduction in Token Consumption] Jinshi Data, May 8 — SenseTime (00020.HK) officially launched its new generation lightweight multimodal intelligent model, SenseTime RiriXin SenseNova 6.7 Flash-Lite. Utilizing a native multimodal architecture, it can "understand" complex webpage layouts, document structures, and financial charts directly, much like humans. This significantly enhances the success rate of complex long-chain tasks such as data analysis, in-depth research, and PPT generation. Additionally, by eliminating the intermediate layer of visual-to-text conversion, the model achieves a leap in intelligent capabilities with fewer parameters, drastically reducing token consumption during inference. In scenarios like information search, token consumption is reduced by 60% compared to pure text-based intelligent models.