At the Frontier AI Model Forum during the 2025 World Internet Conference Wuzhen Summit, Ant Group announced on Saturday that it has deployed a domestically developed AI computing cluster with tens of thousands of cards. The cluster supports both proprietary and mainstream open-source models, achieving over 98% stability in training tasks. Its training and inference performance rivals international top-tier levels and is fully applied in large-model services for security and risk control. By optimizing model parameter efficiency, data utilization, and computational efficiency, Ant Group aims to advance intelligent experience upgrades across its services. [STAR Market Daily, in Chinese]
Related
