Deepseek AI V3, which is a Chinese app, is now a free artificial intelligence application (AI) in the United States of America! And as much as Sam Altman and his family play “calm” and call the new “impressive” model, they are beaten in a game that dominated it until the end of last week.
The United States versus China or Deepseek vs Openai Sweepstakes is a hot race to occupy the head of the artificial intelligence table. But does Sam al -Tamman really do enough for the American market?
Well, if users are, especially from weI feel that an open source technology model using a “favorable” math and was present for two weeks better than ChatgPT -4.0, Altman in trouble.
On January 23, Chatgpt fell for a few hours; Beings on social media flew that “productivity has decreased” as well. At that moment of time, “competition” was just a word in the English dictionary of Openai.
Many have changed since then. The Chinese AI product is located at the top of the mountain in the United States.
Deepseek R1: Fear unknown open source
In late 2022, Openai appeared for the first time in ChatGPT, causing the discontent of the Chinese AI market that was not ready. The developers in the Asian country rushed to make a half baked bot, in an attempt to match the able technology of the technology giant. None of the products they issued, including Baidu, were not enough to meet the requirements and needs of users.
After a little more than three months after December 2022, Dibsic was Be established By Ai Hedge Fund High Flyer, co -founder, co -founder Liang Wenfeng. Initially, it was rooted in the artificial intelligence unit in the high divorce, the start starting in April 2023 moved to focus on large language models and seek to achieve them Artificial general intelligence (AGI) – A level of artificial intelligence capacity that matches or exceeds human intelligence through various tasks.
Very little is known for Winfing. Users usually do not need to know much about “from”. They are more interested in the results, and in comparison side by side, Deepseek models are blowing from water.
Unlike ChatGPT from Openai, Deepsek R1 meets a promise that Sam Altman’s Openai had defended it once; Open Amnesty International.
The storm has not prevented the restrictions of former President Biden on the exports of chips to China in the past three years Deepseek from training its latest models. The United States wants to be the center of everything, and they tried hard to thwart anything that would help China take the initiative.
For all CNBC modern interviewThe CEO of Scale Ai Alexandr Wang claimed that Deepseek has 50,000 NVIDIA H100 chips, which has not provided any guide to his support.
“What we found is that Deepseek … is the best performance, or almost equally with the best American models.”
Wang also said that the disclosure of details would violate Washington’s invitation to prohibit any exports of these advanced materials to China.
While the accurate numbers related to the cost of building R1 or the number of relevant graphics processing units are unknown, analysts at Jefferies indicate that they have only a cost of 5.6 million dollars, much lower than Openai, Meta, and most of the US budgets in the United States to participate in US. Operations.
Just mention NVIDIA in the entire Deepseek Ai madness put the chips makers in trouble; The share price of 17 % on Monday was all we need to see how the American market felt.
Do the United States government fear China a lot? Well, as President of Donald Trump He said BBC yesterday, it is a “call to wake up”.
Openai needs to do a better job
In a post on January 28 on X, Openai Altman’s CEO said that Deepseek’s R1 is a “impressive model”, to provide significant results of low costs to users. He went on to prepare his followers that Openai would launch better models and that the company welcomes competition.
“It is clear that we will present much better models and also it is active legitimate to get a new competitor! We will withdraw some versions.
Deepseek’s R1 is an impressive model, especially about what they can offer for the price.
Obviously, we will present much better models and also it is legitimate to have a new competitor! We will withdraw some versions.
Sam Al -Taman (Sama) January 28, 2025
OpenAi imposes $ 7.50 per million inputs of O1, while Deepseek offers the R1 a 95 % discount – $ 0.14 per million icons.
In software engineering tasks, Deepseek R1 exceeds an Openai, with 49.2 % on SWE-Bench test that was resolved for 48.9 % of Openai. It is 50 times more efficient than leading the American AI models such as Meta’s Llama 3.1 and CLAUDE 3.5, using a much lower computing power to provide comparable or superior results.
Do you blame anyone for forgetting ChatGPT?
Altman document tone is not enough to hide the weaknesses in Openai’s Modus. Unlike Deepseek, Openai Resist Fully embracing open source practices, a position that attract criticism in an industry in which there is nothing more inevitable than transparency and cooperation.
In front of our eyes directly, it is assumed that American companies will always lead the artificial intelligence race through a more intelligent and effective competitor that works outside the Silicon Valley bubble.
If Openai, with GPT 5.0 or whatever the following model, fails to face open source Deepseek technology with something that the market will welcome with open weapons, the latter’s hegemony is no longer predicted, it is spoiler.
