DeepSeek Founder Liang Wenfeng Named One of Nature's Top 10 People Who Shaped Science in 2025

Liang Wenfeng, the 40-year-old founder and CEO of Hangzhou-based AI firm DeepSeek, has been named by the British journal Nature as one of its top 10 "People Who Shaped Science in 2025." Nature described Liang as a "Chinese finance whizz" whose breakthrough AI models drew global attention and prompted debate about the future of artificial intelligence and the tech industry.

The profile highlights the disruption following the January release of DeepSeek-R1, a reasoning model whose performance suggested that the United States might not be as far ahead in AI as commonly assumed. DeepSeek — a spin-off from Liang's High-Flyer Quantitative Fund — attracted intense scrutiny after a market sell-off on 27 January erased nearly US$1 trillion in technology market value, including roughly US$600 billion from Nvidia alone.

Nature noted that DeepSeek's R1 and an earlier model, V3 (released in December), matched the capabilities of comparable systems from leading Western labs while using only a fraction of the training cost. For context, Nature reported that training Meta Platforms' Llama 3 405B model cost more than ten times as much as DeepSeek's reported expenditures.

DeepSeek's decision to publish its model code and training details as open source was singled out as a pivotal move. Researchers worldwide have adapted the models for domain-specific applications, and the release prompted other firms in China and the US to follow suit with open models of their own.

"In many ways, DeepSeek has been hugely influential," said Adina Yakefu, a researcher at AI platform Hugging Face, in Nature's profile.

Despite US restrictions limiting mainland firms' access to advanced AI chips, DeepSeek demonstrated that alternative engineering approaches and training techniques can produce high-performing models. In September, DeepSeek's engineers published the technical details of how R1 was built and trained in Nature, marking R1 as the first major large language model to undergo peer review in the journal and giving researchers a practical "recipe" for training reasoning models.

Nature's feature also highlighted several other notable scientists on its 2025 list. Among them was Du Mengran, a geoscientist from the Chinese Academy of Sciences, whose 2024 dives into the hadal zone uncovered the deepest-known animal ecosystem at the bottom of the Kuril-Kamchatka Trench. Others on the list included immunologists, public-health negotiators, systems biologists, entomologists, neurologists, physicists, and the first patient saved by personalized CRISPR therapy.

DeepSeek's rise raises important questions about transparency, open science and the economics of AI development. Liang's inclusion in Nature's top 10 underscores how technical breakthroughs, publication choices, and market impacts can intersect to reshape scientific and commercial landscapes.

This article is based on reporting by the South China Morning Post and Nature's 2025 "Nature's 10" feature.