China Telecom develops country’s first MoE models trained entirely on Huawei’s AI chips

China Telecom has developed the country's first Mixture of Experts (MoE) AI models, called TeleChat3, trained entirely on Huawei's Ascend 910B chips and MindSpore AI framework. This marks the first public validation of using only Huawei chips to train MoE models, which are known for high performance with fewer computational resources. The models, ranging from 105 billion to trillions of parameters, were trained by China Telecom's Institute of Artificial Intelligence (TeleAI). TeleAI researchers stated that the Huawei technology met the demands of training large-scale MoE models, establishing a full-stack solution for domestic computational ecosystems. This initiative highlights efforts by US-blacklisted Chinese firms like Huawei to develop AI models using domestically designed semiconductors.

Explore how the topics in this article connect to other news stories