首Token延迟降低32.4%!是石科技集群调度大招,国产推理再提效
中文摘要
是石科技发布TokenFleet调度框架,实现首Token延迟降低32.4%,大幅提升国产AI推理效率。
English Summary
Shi Tech's TokenFleet framework reduces time-to-first-token by 32.4%, significantly boosting domestic AI inference efficiency.
Original Excerpt
最新发布TokenFleet调度框架