Domestic PD heterogeneous reasoning has entered the production environment
2026-09-05 08:00Products & Apps🔥 42.2 heat score
1sources
1days unfolding
42.2heat score
4mentions
SummaryAI generated
On September 3, Qujing Technology signed a strategic cooperation agreement with Moore Thread. Based on Qujing Technology’s self-developed domestic PD heterogeneous technology, Moore Thread’s MTT S5000 intelligent computing card, and the MUSA software platform, both parties will build a domestic AI Token factory based on the ATaaS platform. They will also jointly develop and promote a joint solution using Token Pod as the carrier. This solution decouples the Prefill phase and the Decode phase of inference into different resources for execution: Moore Thread’s MTT S5000 will be responsible for input calculations and KV Cache generation in the Prefill phase, while cooperating with high-bandwidth GPUs to handle the Decode phase. The related solution has already been put into production, and under equivalent service standards, its overall cost-performance ratio is superior to international advanced computing solutions, marking the entry of domestic PD heterogeneous inference acceleration into the production environment.
On September 3, Qujing Technology signed a strategic cooperation agreement with Moore Threads. Based on the deep integration of Qujing Technology’s self-developed domestic PD heterogeneous technology with Moore Threads’ MTT S5000 intelligent computing card and MUSA software platform, both parties will continue to build a high-quality domestic AI Token factory based on the ATaaS platform. They will also jointly develop and promote joint solutions using Token Pods as the carrier. The PD separation architecture decouples the computationally intensive Prefill phase from the memory-intensive Decode phase and assigns them to different resources for execution. Moore Threads’ MTT S5000 will be responsible for input calculations and KV cache generation in the Prefill phase, while working together with high-bandwidth GPUs in the Decode phase. The related solutions have already been put into production, and under the same service standards, their overall cost-effectiveness is superior to international advanced computing solutions, marking the entry of domestic PD heterogeneous inference acceleration into the production environment. Both parties will also deepen technical cooperation…