📢 Flash
Huawei Cloud Discontinues Its Own Pangu 2.0 Model Service, Recommends Switching to DeepSeek and GLM
Huawei Cloud announced via the "Beating AI flash news" on November 1 that its self-developed large model services openPangu-2.0-Flash and openPangu-2.0-Pro will cease sales starting November 6 and will be officially retired on November 30; after retirement, all existing users will no longer be able to invoke these two models.
Although both models have been available for only a few months—openPangu-2.0-Flash, with 92 billion parameters, was open-sourced on June 30, and openPangu-2.0-Pro, with up to 505 billion parameters, was open-sourced on July 31—Huawei re-released the full pipeline code for Pangu 2.0’s pre-training, fine-tuning, and reinforcement learning at the end of September.
Regarding the discontinuation plan, Huawei recommends that openPangu-2.0-Flash users migrate to DeepSeek-V4.1-Flash-1017 and openPangu-2.0-Pro users migrate to Zhipu GLM-5.3.
Huawei has confirmed that a new generation of the Pangu model is under planning, but specific model numbers and launch timelines have not yet been disclosed.
Although both models have been available for only a few months—openPangu-2.0-Flash, with 92 billion parameters, was open-sourced on June 30, and openPangu-2.0-Pro, with up to 505 billion parameters, was open-sourced on July 31—Huawei re-released the full pipeline code for Pangu 2.0’s pre-training, fine-tuning, and reinforcement learning at the end of September.
Regarding the discontinuation plan, Huawei recommends that openPangu-2.0-Flash users migrate to DeepSeek-V4.1-Flash-1017 and openPangu-2.0-Pro users migrate to Zhipu GLM-5.3.
Huawei has confirmed that a new generation of the Pangu model is under planning, but specific model numbers and launch timelines have not yet been disclosed.
