A research team that included Huawei Technologies said it used the company’s Ascend 910C chips to complete post-training for DeepSeek-V4-Pro. This suggested domestic hardware is being used for more demanding AI work beyond inference.

A Shenzhen government social media post on June 5 said the team ran DeepSeek’s largest model to date, with 1.6 trillion parameters, on a cluster powered by at least 1,000 Huawei chips and completed full-parameter post-training.

The project involved Huawei, the Shenzhen Loop Area Institute, Harbin Institute of Technology’s Shenzhen campus, and the Shenzhen Research Institute of Big Data.

The post said domestic computing power had previously been used mainly for inference, the process of running an already trained model to answer prompts, while full-parameter post-training updates the model’s entire architecture rather than only selected parts.

The result marks a step in China’s efforts to move from AI inference to model training using domestic hardware under tighter US restrictions.

🔗 Source: South China Morning Post

Recent Huawei developments