Companies of the technological sector and artificial intelligence (AI), such as NVIDIA Y OPENAI they have ruled before him Deepseek generative ouge of the generative the IAunderlining the advance in artificial intelligence which offers, specifically, in reference to your R1 model capable of reasoning.
The Deepseek assistant, developed by the Chinese technology with the same name, is a Free generative ia assistant that, after its launch on January 11 at the App Store, it has managed to occupy the First position in the Apple store in the United States, Ahead of Openai’s ‘chatbot’, Chatgpt, which has been the most popular for more than two years.
Before Deepseek’s boom, sector companies such as Nvidia and OpenAi have underlined their abilities, putting on the table the progress in generative that supposes This assistant, above all, in reference to his R1 reasoning models, despite the fact that his success has caused the NVIDIA itself and other technological firms.
Specifically Nvidia announced on Monday a fall of about 17%, which means a reduction of its capitalization around 500,000 million dollars (almost 480,000 million euros), and which is reflected as the greatest intradicted loss ever registered. The same has happened with AMD (-4,77%), Arm Holdings (-8,4%), Micron (-9,8%) y ASM International (-11%).
This fall is due, in part, to the impact that Deepseek may have on the demand for high -end microchips, since the Chinese firm aims to offer a competitive performance, despite using lower cost chips and less data with an open source model. This allows Reduce the price of use of its AI technology.
Nevertheless, Nvidia It has ruled underlining the family of recently presented reasoning models DeepSeek-R1-Zero y DeepSeek-R1which has described as “a Excellent advance in artificial intelligence “ and an example of “how new models” can be created “using the technique of test time scalingas detailed in a statement collected by Bloomberg.
This scaling technique in trial time is based on the fact that the models can provide better answers if they use additional computer power that allows them Reason when making predictions or generating text and images.
Following this line, the technological one has highlighted how Deepseek has taken advantage of “widely available models and computing that fully complies with export control “of the United Statesin reference to the chips used by the Chinese firm to train its models, which are versions of the NVIDIA GPUs for the Chinese market.
Nvidia also has stressed that Inference in AI “requires a significant number of GPUS NVIDIA and high performance networks”so Deepseek’s advance will continue to drive the products of the American chip manufacturer.
“An impressive model,” according to Openai
For its part, the OpenAi Ejection Director, Sam Altmanhas also ruled on the capabilities of the Deepseek model, which he has defined as “a Impressive model“, as shared in an X (old Twitter) publication.
Specifically, Altman has underlined this reasoning model “especially taking into account what you can offer for the price “, Since, for the moment, the assistant allows to carry out all its functions for free for users, and has been trained with cheaper chips.
Altman has also pointed out that “It’s really stimulating” having a new competitor like Deepseek and that, from OpenAi, they will offer “much better” models in their next releases.
With all this, he has reflected that they will continue to execute their research map and that “now more than ever” they must Have more computer resources To achieve your advance purposes successfully. “We hope to offer you agi and more”, He has sentenced, in reference to the general artificial intelligence in which they are working.
“Creative Solutions” to advance in the field of lower cost
In addition to all this, the ancient CEO of Intel, Pat Gelsinger, has also shared the advances that Deepseek presents For the technological and AI sector, above all, reflecting the ability to offer high performance of AI to a lower cost.
As detailed in an X publication, the abilities offered by Deepseek remember “three important teachings in the history of computer science.” Specifically, Gelsinger has indicated that get lower costs For the development of this technology will allow expanding the AI market.
He also pointed out that engineering “has limitations” and that, as Deepseek has shown, despite having limited resources, Engineers can “find creative solutions“Finally, he stressed that the Chinese firm” will help restore the increasingly closed world of foundational AI models“, thanks to the use of open models.
In this context, Gelsinger is already using Depseek’s R1 model in its own Gloo messaging company. As he has shared in statements to Techcrunch, his team is using this model to develop an AI service to which he refers to Kallmetthat will have a ‘chatbot’ formatamong other options.
Thus, he pointed out that he is using R1 because, although his engineers could have executed the OPENAI O1 model, “you can only access O1 through API”. However, the Depseek model is open source, so they have been able to use it integrally to build its Kallm service.
As Gelsinger has concluded, Depseek can demonstrate that AI advances “through creativity in engineering, without investing more ‘hardware’ power and computer resources in the problem.”