Cambricon's MLU series with the NeuWare stack — a self-reliant domestic option for large-model training and inference.
Together with Ascend and Hygon, a domestic supply matrix.
Domestic silicon route for diverse sovereignty needs.
Unified stack for training and inference.
Schedulable with Ascend and NVIDIA.
Covers training, inference and industry use.
Domestic large-model training, inference and industry deployment.
NeuWare keeps improving with solid mainstream framework support.
Yes, unified heterogeneous scheduling.
Contact us for models and pricing.