Huawei has open-sourced the openPangu-2.0-Pro model — a 505 billion-parameter model — releasing model weights, base inference code and a technical report. The openPangu brand targets Ascend-native training and inference, positioning the release as a best-practice reference to support Ascend adoption and accelerate AI commercialization and ecosystem development. Related components are already live on open-source platforms, and the model is available for online testing via Huawei Cloud MaaS. Huawe

Huawei has open-sourced the openPangu-2.0-Pro model — a 505 billion-parameter model — releasing model weights, base inference code and a technical report. The openPangu brand targets Ascend-native training and inference, positioning the release as a best-practice reference to support Ascend adoption and accelerate AI commercialization and ecosystem development. Related components are already live on open-source platforms, and the model is available for online testing via Huawei Cloud MaaS. Huawei invited global developers, enterprise partners and researchers to download and provide feedback to support the open-source ecosystem.