PANews reported on September 29th that the DeepSeek-V3.2-Exp model was officially released and open-sourced today. The model incorporates a sparse attention architecture, which effectively reduces computing resource consumption and improves model inference efficiency. The model is now available on Huawei Cloud's MaaS platform. Huawei Cloud continues to deploy the DeepSeek-V3.2-Exp model using a large EP parallelization solution, leveraging the sparse attention structure to achieve a long-sequence affinity context parallelization strategy while balancing model latency and throughput performance.