Platform for AI (PAI) - EAS supports concurrent deployment of EP experts
Aug 25 2025
Platform for AI (PAI)EAS supports Expert Parallelism (EP) deployment and provides customers with a high-performance inference deployment solution for MoE models.
Content
Target customers: customers who deploy Mixture of Experts (MoE) services based on EAS. New features /specifications: EAS enables EP deployment of MoE models such as DeepSeek-R1 and supports inference engines such as vLLM and SGLang. This enables customers to break through hardware limitations and improve resource utilization and system throughput.