On August 13, Alibaba made waves in the tech industry by announcing the open-source release of its cutting-edge ultra-large-scale Mixture of Experts (MoE) model, named Qwen3.8-2.4T-A95B. In a parallel development, the Zhongzhi FlagOS community has successfully accomplished the adaptation of this model across a diverse array of nine AI chips. These chips hail from prominent manufacturers including T-Head, NVIDIA, Moore Threads, Huawei Ascend, MetaX, Kunlunxin, Hygon, Qingwei Intelligence, and Enflame.

This remarkable feat was made possible through the utilization of the FlagOS unified open-source technology stack. This robust framework facilitated seamless multi-chip adaptation, ensuring precision alignment and rigorous deployment verification. The FlagOS technology stack, known for its versatility and efficiency, has played a pivotal role in bridging the gap between advanced AI models and a wide range of hardware platforms.