WM32 integrates cutting-edge heterogeneous multi-core NPU V3 architecture, offering unparalleled performance and efficiency. It supports low-bitwidth mixed precision and features high-performance tensor cores specifically optimized for Convolutional Neural Networks (CNNs), as well as accelerated computation for Transformer and BEV applications. With flexible vector cores that are designed to enhance Transformer networks, the WM32 delivers outstanding performance, supported by SDMA and seamless multi-core hardware coordination.