China's AI model, DeepSeek, has taken the tech world by storm, surpassing ChatGPT, Gemini, and Claude AI. The chatbot has climbed to the top of Apple's App Store rankings in the US, dethroning ChatGPT and shaking up the stock market. This rise is led by a team of young engineers working with limited resources but remarkable expertise.
Luo Fuli: China's AI Prodigy
One of DeepSeek's key figures is Luo Fuli, a 29-year-old AI researcher. Despite struggling with computer science in her early years at Beijing Normal University, Luo excelled at Peking University's Institute of Computational Linguistics. In 2019, she published eight research papers at the ACL conference, gaining the attention of top tech firms.
![Luo Fuli Luo Fuli](https://data.ibtimes.sg/en/full/79548/luo-fuli.jpg?w=736)
She later joined Alibaba's DAMO Academy, contributing to projects like VECO and AliceMind. In 2022, she moved to DeepSeek, where her expertise in natural language processing helped develop DeepSeek-V2, a model that rivals ChatGPT. Her growing reputation attracted Xiaomi's founder, Lei Jun, who reportedly offered her an annual salary of 10 million yuan. However, reports suggest she has not accepted the offer.
DeepSeek: China's AI Powerhouse
DeepSeek was founded in 2023 by Liang Wenfeng in Hangzhou. The company has developed powerful large language models (LLMs) at a fraction of the cost of global competitors. DeepSeek-V3 has been benchmarked against models like Llama 3.1 and Qwen 2.5, performing on par with OpenAI's GPT-4o and Anthropic's Claude 3.5 Sonnet.
According to DeepSeek, its R1 model outperforms OpenAI's o1-mini model in multiple benchmarks. Research by Artificial Analysis also indicates that DeepSeek-R1 surpasses AI models from Google, Meta, and Anthropic in quality.
DeepSeek's chatbot has quickly risen to the top of the US App Store, becoming the most downloaded free app. This success has disrupted the AI industry, leading to declines in the stock prices of Nvidia, Microsoft, and Meta.
Breaking Barriers Despite US Restrictions
DeepSeek's success is even more impressive given US-imposed restrictions on China's access to advanced semiconductors. Despite these challenges, the company has managed to develop high-performance AI models using fewer resources than its Western counterparts.
Industry leaders have taken notice. OpenAI CEO Sam Altman acknowledged DeepSeek's achievements, saying, "DeepSeek's R1 is an impressive model, particularly around what they're able to deliver for the price."
The Visionary Behind DeepSeek
DeepSeek's founder, Liang Wenfeng, rarely gives interviews. However, in discussions with Chinese tech media outlet 36Kr, he emphasized his vision for AI. Liang believes China must be a global innovator rather than just a consumer of technology.
He has criticized Chinese firms for focusing too much on business applications instead of cutting-edge AI research. DeepSeek, by contrast, has pursued the most complex AI challenges, leading to the success of its latest models.
Liang, born in 1985 in Guangdong province, studied information engineering at Zhejiang University. He first gained recognition in AI through High-Flyer Quant, a quantitative hedge fund he founded in 2015, which managed over 100 billion yuan at its peak.
DeepSeek's Young and Talented Team
DeepSeek's research team includes young scientists from China's top universities, such as Tsinghua and Peking University. The company favors fresh graduates but also hires PhD candidates and experienced AI professionals.
The technical paper for DeepSeek-R1 credits around 200 researchers, with 18 key contributors playing central roles. The team's innovation and agility have driven DeepSeek's rapid rise in AI.
A Unique Business Model
Unlike many Chinese AI startups, DeepSeek operates with just 10 million yuan in registered capital. However, it benefits from High-Flyer Quant's resources, securing over 10,000 Nvidia GPUs despite US semiconductor export restrictions.
DeepSeek has also avoided external funding, diverging from competitors like Zhipu AI and Moonshot AI, which raised over 20 billion yuan in 2024. Instead, DeepSeek follows an unconventional structure, allowing employees to define their own roles without rigid hierarchies.
This flexible approach has enabled DeepSeek to challenge tech giants, proving that innovation can thrive even with limited resources.