FT announced the postponement of the release of the new DeepSeek model due to problems with chips

DeepSeek, a Chinese artificial intelligence (AI) company, has postponed the release of a new chatbot model due to problems with learning on Huawei chips. This was reported on August 14 by the Financial Times (FT) newspaper.

According to the FT, the Chinese government recommended that developers use domestic Huawei Ascend chips instead of American Nvidia processors. As a result, DeepSeek faced technical difficulties in developing a new model of its R2 chatbot.

It is clarified that AI developers had to use Nvidia chips at the stage of training the neural network, and during testing of the finished model for generating responses, they had to return to Huawei processors. The newspaper's sources stressed that the release of the new model, despite all the difficulties, could take place in the coming weeks. Work to improve the compatibility of Huawei chips and neural network models is also continuing, the FT concluded.

Go to GPT: how is the new version of the neural network better than the previous models

The effectiveness of GPT-5 was compared with a team of experts

Bloomberg reported on August 12 that Chinese authorities urged local companies to refrain from using Nvidia's H20 processors, especially in the public sector. There is no direct ban on the use of H20 in China. Local technology companies want to get these processors that work well in artificial intelligence (AI) applications. However, the government's calls create difficulties for Nvidia, which is trying to compensate for losses from restrictions on chip sales to China.

All important news is on the Izvestia channel in the MAX messenger.