Huawei reportedly hit back at suggestions its AI research lab used elements from rival models to develop its own Pangu Pro platform, stating it respected intellectual property and stuck close to licensing terms.

In a post on WeChat, seen by Bloomberg, the Chinese vendor responded to accusations posted on coding website GitHub which claimed the Pangu Pro Mixture of Experts (MoE) model had used uncredited source code from rivals.

The accusation came after Huawei released its source code for the model last week, prompting a group to be set up on GitHub called “HonestAGI’, calling out elements of Alibaba’s Qwen 2.5 14B model in particular.

Noah’s Ark research lab, the team behind Pangu Pro MoE, stated in response it strictly adheres to the requirement of open source licenses “and clearly mark copyright statements in the relevant source files”.

It continued: “We welcome and look forward to in-depth and professional discussions on technical details with everyone in the open source community.”

The research lab was founded in 2012 and introduced a number of Pangu AI models in 2021 for various sectors, including telecoms and weather forecasting. The Pangu Pro MoE has 72 billion parameters and uses a hybrid technique powered Huawei’s in-house designed AI chips.