vLLM Ascend Plugin
A community-maintained hardware plugin that runs the vLLM inference engine on Huawei Ascend NPUs.
vLLM Ascend is a hardware plugin that lets the vLLM inference engine run on Huawei Ascend NPU hardware, following vLLM's hardware-pluggable architecture so Ascend support integrates without changes to vLLM's core. It is the community's recommended path for running large language models on Ascend hardware through vLLM.
The project publishes official versioned releases alongside vLLM's own release cadence, documents deployment guidance for scenarios such as large-scale expert-parallel serving, and highlights user stories describing fine-tuning, evaluation, reinforcement learning, and deployment workflows built on it. It is developed in the open with the wider vLLM community, including joint meetups and shared release announcements, and is licensed under Apache 2.0.