DeepSeek: A Chinese LLM that can solve some problems at the same standard in o1, DALLE 3 and Stable Diffusion
Some auto industry watchers felt a creeping sense of repetition. Seemingly out of nowhere, a Chinese firm made international headlines by besting Western companies at the tech they supposedly invented.
No, it wasn’t BYD, the 20-year-old automaker that gained sudden global recognition in recent years as it began to export low-price electric vehicles all over the world. (BYD built more electric vehicles in 2024 than Tesla.) The buzz of the week was about a Chinese startup that shocked techies when it released a new open-sourced artificial intelligence model with a fraction of the funding US competitors have gotten to build their own. US tech stocks fell earlier this week, and investors were scrambling to reexamine their bets.
Many of China’s major global economic success stories have emerged out of a similar national strategy, says Susan Helper, an economist with Case Western Reserve University who studies global supply chains and manufacturing and worked on EV policy in the Biden administration. She says it’s basically a decision on an industry that’s critical and put a lot of money towards it for a long time. The US approach to cars is similar to where we change our minds on electric vehicles every few years.
Chinese car makers have been subsidized for nearly two decades, and the government has given tax breaks to people who use electric vehicles, in order to push the country to go electric. Chinese investments of artificial intelligence are growing bigger. The Chinese government has invested more than $200 billion in artificial intelligence-related companies over the past decade, according to researchers. Just this month, it announced a new $8.2 billion AI investment fund.
It was only a matter of time before DeepSeek would emerge in China, because of the huge investment in firms by venture capital that develop degrees in science, technology, engineering and mathematics. There would be other Chinese LLM that could do great things if there was no DeepSeek.
On 20 January, the Hangzhou-based company released DeepSeek-R1, a partly open-source ‘reasoning’ model that can solve some scientific problems at a similar standard to o1, OpenAI’s most advanced LLM, which the company based in San Francisco, California, unveiled late last year. Openai’s DALLE 3 and Stable Diffusion were both created by Stabilityai in London, which is why Deep Seek launched a model called Janus-Pro-7B this week.
In fact, there are. On 29 January, tech behemoth Alibaba released its most advanced LLM so far, Qwen2.5-Max, which the company says outperforms DeepSeek’s V3, another LLM the firm released in December. And last week, Moonshot AI and ByteDance released new reasoning models, Kimi 1.5 and 1.5-pro, which the companies claim can outperform o1 on some benchmark tests.
It was a priority to develop a talent line of ‘AI talent’. The Chinese ministry of education gave the go-ahead for over 500 universities to offer undergraduate degrees in Artificial Intelligence by 2022, according to a Georgetown University report. The United States accounted for only 18% of the world’s leading artificial intelligence researchers in that year.
Wenfeng, at 39, is himself a young entrepreneur and graduated in computer science from Zhejiang University, a top institution in Hangzhou. He co-founded the hedge fund High-Flyer almost a decade ago and established DeepSeek in 2023.
Jacob Feldgoise, who studies AI talent in China at the CSET, says national policies that promote a model development ecosystem for AI will have helped companies such as DeepSeek, in terms of attracting both funding and talent.
Despite the rise of artificial intelligence courses, it is not clear how many students graduating with degrees from universities that teach the skills companies need. Chinese AI companies have complained in recent years that “graduates from these programmes were not up to the quality they were hoping for”, he says, leading some firms to partner with universities.
