DeepSeek – ChatGPT analogue or its "killer"?

[bitheerani319]

In January of this year, a Hangzhou-based Chinese company specializing in the development of artificial intelligence (AI) unveiled the R1 version of an AI model called DeepSeek . The news literally shook the world of the tech industry, which was most clearly manifested in a sharp drop in the shares of graphic processing unit ( GPU) manufacturers on the stock markets, something that had never been recorded before in the United States. What is the phenomenon of the “Chinese miracle” and how serious could the consequences be?

What is the secret of DeepSeek's success?
The AI ​​model presented by Chinese engineers is a variation of the large language model LLM ( Large Language Model) , which is used to usa whatsapp data neural networks for AI applications. On their basis, it is possible to build AI systems for implementation in any field of knowledge - IT, automation, linguistics, finance, etc.

The fact that another AI model has appeared is not in itself a reason for excitement, because hundreds of them appear in the world every year and usually only narrow specialists pay attention to this. But not this time.

The first version of the language model, called DeepSeek Coder, was released quite recently – in November 2023 and was focused on solving programming problems. At that time, few people paid attention to it. This was followed by versions VL (March 2024) and V2 in May of the same year. The latter version was so successful that it caused a collapse in prices on the market of Chinese brands working in the IT field – Alibaba, Baidu, Tencent and many others.

In November last year , DeepSeek V3 was released , which made a breakthrough in the speed of answering compared to earlier versions. The January 2025 version of DeepSeek R1 was built on the basis of V3 and absorbed its best features. Its main task was to solve “logic” problems and perform mathematical calculations in real time.