DeepSeek-V3.2-Exp Model Officially Released and Open-Sourced

The DeepSeek-V3.2-Exp model has been officially released and open-sourced today. The model introduces a sparse Attention architecture, which effectively reduces computational resource consumption and enhances inference efficiency. Currently, the model has been officially launched on Huawei Cloud's large model as a service platform (MaaS). For the DeepSeek-V3.2-Exp model, Huawei Cloud continues to adopt the large EP parallel deployment solution, leveraging the sparse Attention structure to implement a long-sequence-friendly contextual parallel strategy while balancing model latency and throughput performance.