DeepSeek has officially open-sourced infrastructure components for Huawei's Ascend computing platform, including the TileLang high-level language compiler, compute libraries, and distributed communication libraries. According to ChainCatcher, the release mirrors the open-source components previously made available for Nvidia platforms.

TileLang is designed to provide a general-purpose, easier-to-program high-level language that can reach hardware performance limits, while improving development efficiency and simplifying logic compared with CUDA. DeepSeek said the Ascend version wraps the underlying Ascend C instruction set to offer high-level programming without sacrificing hardware performance.

The company also open-sourced DeepGEMM, DeepEP, TileKernels, FlashMLA, and DeepSelect, which are used for general matrix operations, large-scale cross-device communication, standard vector computation and memory access, long-context sparse attention, and data filtering. DeepSeek said related compute and communication performance has approached hardware limits in multiple key tests, and that Huawei's team supported development of a 128-card supernode solution based on Ascend 950.