One thing that dominated China-related discourse this weekend was DeepSeek, a “dark horse” AI model that seemed to have popped up out of nowhere, causing a “Sputnik moment”. The Chinese-made app went from topping the free apps list on Apple’s App Store in both China and the United States to featuring on Google Trend as one of the most searched-for terms, all the way to causing GPU-maker Nvidia’s share prices to drop 14% on Monday. Meta has reportedly put together 4 research groups to analyse the model to improve its own Llama large language model (LLM). But what is it, and what are the Chinese people saying about it?
Before the current version (R1) was launched in late January and broke the internet, the V3 version, released less than a month ago, already had AI-related professionals and pundits calling it the successor to OpenAI. The R1 and DeepSeek’s essays on the model are called “gems” not only because of the AI’s cost-effectiveness—it achieved similar performance to Open AI’s GPT-o1 with a tenth of its cost—but also because they are open source.
On Weibo, China’s Twitter-like platform, many DeepSeek-related topics made it to the Hot Search list over the weekend. The hashtag “DeepSeek” itself topped the list with 110 million views. Interestingly, topics such as the DeepSeek founder returning to Guangdong for the Chinese New Year (CNY) and that the team is largely made up of new graduates reached numbers 4 and 9, with 41.87 and 29.11 million views, respectively. Most of the comments are proud of the younger developers and their achievements, as some believe the model would break the “American hegemony” in AI models. Others argue that DeepSeek might not be the “innovation with Chinese characteristics” they’ve been waiting for as their secret ingredient is still “Silicon Valley” flavoured.
The market, of course, quickly picked up on the viral popularity of the model. While the Nasdaq and Dow tumble, China’s Shanghai and Shenzhen are reacting with optimism. A collective of Chinese tech or related stocks called “DeepSeek concept stock” went up to 10% or reached limit up on Monday morning. However, some netizens find DeepSeek’s, especially its founder Liang Wenfeng’s involvement with quantitative hedge fund High-Flyer, evokes suspicion around the boom right before the market closes for CNY. With researchers and competitors trying to recreate the model with its open-source resources, we are likely to find out more about this new model very soon.
