Huawei-led team claims it post-trained DeepSeek's 1.6-trillion-parameter model — 1,000 Ascend 910C chips used in training - Tom's Hardware

Huawei has announced that its research team has successfully post-trained the DeepSeek model, which features a staggering 1.6 trillion parameters. This development is significant as it marks a notable advancement in AI model capabilities, particularly in the context of the increasing demand for more powerful and efficient machine learning systems.

The training of DeepSeek utilized an impressive array of 1,000 Ascend 910C chips, showcasing Huawei’s commitment to pushing the boundaries of AI technology. The Ascend 910C is part of Huawei’s proprietary AI chip lineup, designed to enhance computational efficiency and performance. This achievement not only highlights Huawei’s technical prowess but also positions the company as a formidable player in the AI landscape, competing against other tech giants like NVIDIA and Google, which have also been investing heavily in AI infrastructure.

The implications of this development are substantial. With 1.6 trillion parameters, DeepSeek could potentially outperform existing models in various applications, from natural language processing to complex data analysis. This leap in model size and capability could lead to more sophisticated AI applications, enabling businesses to derive deeper insights and automate processes more effectively. As noted by Tom’s Hardware, the scale of this model could redefine benchmarks for AI performance, challenging existing paradigms and pushing competitors to accelerate their own innovations.

As the AI landscape evolves, it will be crucial to monitor how DeepSeek’s capabilities are leveraged across industries and whether Huawei can maintain its competitive edge in the rapidly advancing field of artificial intelligence.