On September 19, Zhu Zhaosheng, head of Huawei's computing strategy and business development department, said that Shengteng has crossed an ecological inflection point and is fully capable and has the opportunity to build a new AI ecosystem in the Agentic era. From a community perspective, the average number of active monthly users in the Shengteng CANN open source community exceeds 5,200. Since June of this year, it has become the most active open source community in China. The number of non-Huawei developers in the community exceeds the number of Huawei developers. Currently, more than 30,000 lines of merged code are added every day. At the model level, there are more than 40 large language models and multi-modal models that have been pre-trained based on Shengteng. Shengteng has become the only AI computing power platform in China that supports pre-training. At the ecological level, in addition to its own open source CANN, Shengteng has connected to mainstream open source frameworks such as PyTorch.

Zhitongcaijing · 1d ago
On September 19, Zhu Zhaosheng, head of Huawei's computing strategy and business development department, said that Shengteng has crossed an ecological inflection point and is fully capable and has the opportunity to build a new AI ecosystem in the Agentic era. From a community perspective, the average number of active monthly users in the Shengteng CANN open source community exceeds 5,200. Since June of this year, it has become the most active open source community in China. The number of non-Huawei developers in the community exceeds the number of Huawei developers. Currently, more than 30,000 lines of merged code are added every day. At the model level, there are more than 40 large language models and multi-modal models that have been pre-trained based on Shengteng. Shengteng has become the only AI computing power platform in China that supports pre-training. At the ecological level, in addition to its own open source CANN, Shengteng has connected to mainstream open source frameworks such as PyTorch.