DeepSeek-R2 Release Allegedly Postponed Amidst Huawei Chip Challenges
2025-08-14 / Read about 0 minute
Author:小编   

The Chinese AI startup DeepSeek initially anticipated launching its latest AI model, DeepSeek-R2, in May but has reportedly pushed back the release to an undisclosed date between August 15th and 30th. DeepSeek-R2, which is powered by a cluster of Huawei Ascend 910B chips, is expected to boast a total of 1.2 trillion parameters, nearly doubling the capacity of its predecessor, DeepSeek-R1. Although the news of a delay due to training issues with Huawei chips has not been officially verified, DeepSeek-R2 is highly anticipated within the industry for its potential to drastically reduce training and inference costs compared to GPT-4, alongside its exceptional performance.