The Chinese launched a spectacular syhetic ielligence model yesterday; Deepseek; This new artificial ielligence not only challenged previous notions about artificial ielligence, but some analysts believe it was also effective in falling the crypto market.
Dip Sik’s supply by Chinese companies eliminated the misconception that China could only produce limited artificial ielligence models, showing that Openai had an unstable advaage, as well as previous reports that China has been in the developme of artificial ielligence six mohs from the states. The United States is back in question.
What is Deepseek and why did it be released?
The large DEEPSEK-V3 language model was unveiled late in the year, and the Chinese claimed that they were built in only six mohs with a budget of $ 1.2 million, compared to the astronomical budgets of artificial ielligence companies based in Silicon Valley.
The United States applied a sanction to China even on October 1 (October 1). These sanctions preveed the export of special computing chips from US companies such as NVIDIA and AMD.
Despite the ban, Chinese companies turned to domestic production or used cloud computing platforms such as Amazon Cloud Services to bypass restrictions and achieve these advanced chips.
China also facilitated its artificial ielligence developme regulations to encourage innovation in August. The coury, for example, canceled financial penalties for companies that had violated industrial regulations.
In order to understand the importance of Deepseek’s budget, we must keep in mind that the cost of building and developing GPT chat will reach $ 5 billion. However, the cost of designing and producing Deep Sik was less than $ 5 million.
However, users’ reports on social networks also show that Deep Six is also better than the GPT chat, and the protocol has newer data.
Deepseek also released a newer model called Deepseek-R1 a few days ago on January 5. In various tests, the model was able to trunk to the trunks such as “GPT-4O” and “Claude Sonnet 3.5 Ahropic” and even better than the famous models such as Llama 3.1 Meta and Alibaba’s QWen2.5.
It is said that the “R1” model is now surpassed in many experimes from the O1 model. This dramatic performance with a fraction of the cost of big competitors and, of course, has surprised artificial ielligence experts. Some have even believed that Chinese artificial ielligence models will gradually surpass their American couerparts.
