However, in overseas social media and even on Wall Street, the popularity of DeepSeek-R1 has just begun to spiral up.

On January 24th, in the foreign big model ranking Arena, the benchmark test of DeepSeek-R1 has risen to the third place in the whole category of big models, among which the StyleCtrl category ranks first with OpenAI O1. However, its arena score reached 1357 points, slightly exceeding the 1352 points of OpenAI O1.

This is another DeepSeek-V3 attack on the world's strongest AI model after DeepSeek-R1 ranked first in the open source model category after eliminating closed source models such as OpenAI o1.

Just after the Arena was released, the global discussion about DeepSeek escalated again. Amazing, beautiful and even conspiracy theories emerge one after another. At the same time, the carnival of AI economy in the U.S. stock market also began to vibrate under the influence of DeepSeek.

US Stock alarm

just after DeepSeek-R1 won the first place in style control model scores, Yann lecun, chief scientist of Meta AI, who rarely speaks, said on social media on the 24th: "When people see DeepSeek's performance, exclaim that" Chinese AI is catching up with America ", but this interpretation is wrong. The more accurate conclusion is that the open source model is surpassing the closed source system. DeepSeek's success benefits from the open ecology, and its innovation is based on the continuous iteration of previous achievements.

The release of DeepSeek-R1 is like a heavy stone hitting the lake of AI industry and AI investment, and even starting to alert more and more people to the possible bubble of AI investment.

THE SHORT BEAR, a US stock giant, said on social media that DeepSeek created a painful moment for AI giants, and investors must sound THE alarm bell for this.

"If the amount needed to defeat OpenAI is 55 million US dollars (including 5,000 H800 GPU and 5 million pre-training fees), the commercialization of this industry will be much faster than many people expected.