Huawei Opensources AI Models Pangu 7B 72B to Boost Global Ecosystem

Huawei has announced a significant move in the AI industry by open-sourcing two of its Pange models and model inference technologies based on Ascend. This initiative is aimed at building its AI ecosystem and expanding its presence overseas. The tech giant revealed the open-sourcing of its Pangu dense model with 7 billion parameters, the Pangu Pro MoE (Mixtures-of-Experts) model with 72 billion parameters, and its model inference technology based on Ascend, which serves as the platform for AI infrastructure. The weights of the Pangu Pro MoW 72B model and the large-scale MoE model inference code based on Ascend have been open-sourced, while the inference code for the Pangy 7B will be launched on the open-source platform soon.

This move aligns with other Chinese AI players pushing an open-source development strategy. The firm argued that it would help speed up the adoption of artificial intelligence across multiple industries globally. According to the tech company, its Ascend ecosystem refers to AI products developed around its Ascend AI chip series. The firm acknowledged that the chips are widely considered to be China’s leading competitor to AI products from Nvidia. The U.S. multinational tech corporation is limited from selling its advanced products to China.

Lian Jye Su, chief analyst at Omdia, argued that Pangu advancing to open-source allows developers and organizations to test the models and add personalized customizations for their needs. He also believes the initiative will incentivize the use of other Huawei products. Su acknowledged that coupling Huawei’s Pangu models with its AI chips and related products gives the firm an advantage in optimizing its AI solutions and applications. Huawei has focused on specialized AI models in sectors like government, finance, and manufacturing, despite competitors like Baidu having LLMs with broad capacity.

Huawei’s latest announcement also aligns the company with its international ambitions, where it has been slowly integrating into new overseas markets. The Chinese tech company also invited developers, corporate partners, and researchers globally to download and use its open-source products to gather and improve feedback. Marc Einstein, Research Director at Counterpoint Research, believes Huawei’s open-source will resonate well in developing countries where companies are more price-sensitive, as with its other products. Huawei also wants to bring its latest AI data center solutions to new countries as part of its global strategy.

Huawei’s open-sourcing move is part of a broader strategy to transform from a competent private sector telecommunications company into a tech giant straddling the entire AI hardware and software stack. This initiative is another key measure of its Ascend ecosystem strategy. The firm’s objective is to ultimately use open-source products to drive hardware sales, which is a completely different model from others. It also collaborates with other tech companies and will continue to do so.

Principal analyst at Constellation Research Ray Wang compared Huawei’s chip-to-model strategy to that of Google, which is also developing AI chips and AI models like its open-source Gemma models. This comparison highlights Huawei’s strategic approach to integrating its AI hardware and software solutions, aiming to optimize its AI applications and solutions. The firm’s focus on specialized AI models in sectors like government, finance, and manufacturing, despite competitors having LLMs with broad capacity, shows its commitment to tailored AI solutions.