Huawei has announced open access to AI models from the Pangu series. The company has already released the weights of the Pangu Pro MoE model with 72 billion parameters, as well as the source code for basic deployment and auxiliary tools for working with large MoE models. All these materials are available on the GitCode platform.
Pangu Pro MoE is designed to handle a wide range of natural language processing tasks. Huawei has also provided model execution technology optimized for its own Ascend chips. This allows for efficient running of large language models on the company’s hardware.
Among other announced products is the Pangu 7B language model, which contains 7 billion parameters. Huawei has stated that its release will happen soon. Users will be able to apply these models for text generation, translation automation, and other tasks related to text information analysis.
The opening of access to Pangu Pro MoE and the launch of Pangu 7B follows a similar move by Baidu, which recently open-sourced the Ernie 4.5 model.
