[Xiwang Releases Next-Generation Inference GPU Chip S3] Xiwang today unveiled its next-generation inference GPU chip, QiWang S3, along with the launch of the HuanWang SC3-256 supernode solution designed for large-scale model inference. This solution supports multi-modal MoE inference scenarios with hundreds of billions to trillions of parameters. Additionally, Xiwang has partnered with ecosystem players such as SenseTime and Fourth Paradigm to jointly initiate the 'One Million Tokens for One Cent' inference cost plan. Xiwang also signed a strategic cooperation agreement with Zhejiang University to establish the 'Intelligent Computing Joint Research Center.' Xiwang, formerly the large chip division of SenseTime, recently announced the completion of nearly 3 billion RMB in strategic financing within a year. (Sci-Tech Innovation Board Daily)