Liang Wenfeng, CEO of Chinese AI startup DeepSeek, is on Nature’s 2025 “Nature’s 10” list. He is recognized for his open-source AI models that challenge U.S. dominance.
DeepSeek launched the R1 model in January. It outperformed GPT-4 in most reasoning tests. In December, DeepSeek released V3, a 405-billion-parameter model.
It matches U.S. systems in coding and reasoning but costs far less. Over 1.2 million downloads followed. Researchers built thousands of derivative projects for drug discovery and climate studies.
The impact reached markets. DeepSeek’s breakthroughs caused a $1 trillion tech stock shake-up. U.S. firms opened their models in response. Liang’s finance background helped him scale DeepSeek efficiently, despite U.S. hardware restrictions.
Researchers use DeepSeek models for protein folding, seismic prediction, and more. Accuracy improved, and research timelines shortened. Monthly usage reached 45 million sessions, with most users outside English-speaking countries.
Liang shares the list with nine other leaders, including Chinese geoscientist Du Mengran. Nature says Liang “redefined AI as a shared scaffold for discovery.” His work shows the global rise of open, collaborative AI.
