Huawei Unveils Industrial Large Language Models

Chinese telco giant Huawei has unveiled the latest in its Pangu line of large language models.

Pangu 3.0, introduced at the company’s Developer Conference, is a series of pre-trained models that can be applied to industrial applications when combined with fine-tuned, industry-specific datasets.

Huawei Cloud CEO Zhang Ping'an said in a statement that the new models will “empower everyone from every industry with an intelligent assistant, making them more productive and efficient.”

The new Pangu model has already been tested in 10 areas including finance, manufacturing, government affairs and health care, according to China Daily.

Huawei said the first commercial use of Pangu will be in the mining operations of Shandong Energy Group. The Pangu Mining Model will be deployed in a mining site in Jinan, the capital of Shandong province in Eastern China. It will be used as an AI-based intelligent monitoring system in the transport of large coal blocks and anchors.

Another potential application of Pangu 3.0 could see the model power weather forecasting. Huawei has published a research paper showing that the model offers a 10,000-times improvement in weather prediction speed to mere seconds.

In 2021, Huawei had claimed that a prior iteration of Pangu was the world’s largest Chinese language model, at 207 billion parameters. This title was challenged by the team that developed WuDao 2.0, which claims to be the world’s largest language model with 1.75 trillion parameters.

At the conference, Huawei also unveiled its Ascend AI cloud services.

This article first appeared on IoT World Today's sister site, AI Business.