梁文锋Liang Wenfeng

Born 1985 in Zhanjiang, Guangdong, China. Father was a school teacher who stressed the importance of learning. Earned a Bachelor's in Electronic Information Engineering from Zhejiang University (2007) and a Master's in Engineering (2010), researching target-tracking algorithms. Applied machine learning to quantitative finance: founded the hedge fund High-Flyer in 2015, which grew to over 100 billion yuan (~$14B) in assets by 2021. Through High-Flyer, built massive computing infrastructure that became the foundation for DeepSeek, launched in May 2023 with the goal of pursuing AGI. DeepSeek's R1 reasoning model (January 2025) shocked the world -- trained for ~$6M vs GPT-4's ~$100M, it matched frontier model performance and caused a brief $1 trillion market cap loss across US tech stocks. DeepSeek demonstrated that Chinese AI could compete at the frontier with a fraction of the resources. On 1 Jan 2026 the team published a new training method that analysts called a 'breakthrough' for scaling, and on 24 Apr 2026 open-sourced the DeepSeek-V4 preview (V4-Pro at 1.6T parameters and V4-Flash at 284B, both with 1M-token context), which absorbed the long-rumored 'R2' reasoning line into a single hybrid-reasoning V4 series.