< IMG SRC = "/Uploads/Blogs/C0/AF/IB-FQLBHV0O_AA088C38.jpg" Alt = "Deepeseek claims that theoretical cost and profit ratio is 545% on the day"/~ 62 < P > Chinese Shi Startap DeEEEK on Saturday, March 1, published some costs about expenses and income, pro-& rsquo; iced with their popular V3 and R1 models. He claims that the theoretical cost ratio of and profit is up to 545% per day.
< p > reports Reuters.
< P > What is known about the costs and income of Deeseek
< p >The Agency notes that this is the first time deepeseek has released information about its profit at the & laquo; withdrawal & raquo; & mdash; Stages after training model, when AI systems make predictions or perform tasks, for example, provide chatbots.< P > According to Reuters Sows, this discovery can further undermine shares of artificial intelligence outside China, which sharply fell in January after chatbots on the Internet and applications based on R1 and V3 models sharply gained popularity around the world. < p > pre -& nbsp; deeppesek & nbsp; stated that it spent about $ 6 million on chips used for modeling the model. It's much less than spent competitors with the US such as Openai.
< p > A news agency emphasizes that H800 chips from NVIDIA used by Deepeseek are also much less powerful than those that Openai and other US firms have access to. It fueled more about doubts about the huge costs of the most up-to-date equipment in the American Shi-sector.
< P > Deeseek notes that if the rent of one chip H800 is $ 2 per hour, and the total daily value of conclusions for its models V3 and R1 & nbsp; & mdash; $ 87 072. & nbsp;
< p > in turn theoretical daily income from these models is $ 562 027, which leads to costs and profit 545%. This year will add over $ 200 million income.
< P > However deepeseek added that its & laquo; actual income is much lower & raquo;, since the cost of using the model V3 is lower than R1.
~