How did Deepsek become so cheap? Doubt started getting up, the report said- misleading

agency:AyraNews24x7W

Last update:January 29, 2025, 15:11

It is being claimed that in making Deepsek, it has cost less than Chatgpt. The report said that the deep AI Assistant has been made for $ 5 million, which looks misleading.

Experts are looking at the expenses of the expenses in making Deepsek

New Delhi. You will remember that when the chatgpt of OpenAI came, there was a stir in the whole world. Now some similar stir is also created in China, the new AI Deepsek is also creating. This AI assistant has come like a storm in the technical world and there has been a ruckus in the markets around the world. It is being said about Deepseek that it has been prepared at very low cost. This claim has launched a debate among the tech experts from all over the world. People are now questioning the expenses incurred in making it with its ability.

Bernstein has said in a report that the Deepcic is fantastic to watch, but it is not a miracle and has not been made in $ 5 million. The report said that this AI has been constructed in $ 5 million, which is very misleading and this does not clear the entire picture.

Also read- Chatgpt vs. Deepsek: Chinese model challenged the American company, what difference between the two? Who is better?

What does the report say
Bernstein’s report states that we believe that Deepcik has not constructed it for $ 5 million. This AI model looks great, but we don’t think they are miracles. The panic that happened on Twitter in the weekend seems more or unnecessary than necessary.

The report said that the Deepcik has created two families of the AI ​​model- one Deepcic-V3 and the other Deepcic R1. The V3 model is a large language model that uses mixture-of-experts (MoE) architecture. To trend the V3 model, Dipcik used a cluster of 2,048 Nvidia H800 GPU for about two months, including about 2.7 million GPU hours for pre-training and 2.8 million GPU hours after training. Those who do not know, tell them that GPU hour is a total time that takes a graphics processing unit to do processing tasks.

Let us tell you that with the launch of the deputycick, AI Chip Leader Nvidia suffered the most loss and on 27 January there has been a loss of $ 593 billion in the market value, which is a one -day record deficit for any company. However, in what Bernstein concluded, he has admitted that although the achievements of the deputy are fantastic, it is very strong. But only his claim to make OpenAI competitive for $ 5 million seems to be exaggerated.