Textbook of AI

1985–, Quantitative-finance executive turned AI founder

Liang Wenfeng (梁文锋) is a Chinese entrepreneur who in 2023 founded DeepSeek out of the quantitative-investment firm High-Flyer that he had previously co-founded. DeepSeek released a sequence of capable open-weights language models culminating in DeepSeek-V3 (December 2024) and DeepSeek-R1 (January 2025). The R1 model is a reasoning model trained with large-scale pure reinforcement learning that matches or exceeds OpenAI's o1 on many benchmarks while being released under an open licence with full training details.

The DeepSeek-R1 release in January 2025 prompted what was widely called the "DeepSeek moment", a sharp public reassessment of US AI export-control policy, of frontier-lab valuations, and of the competitive landscape between US and Chinese AI organisations. The release demonstrated that frontier-quality reasoning models could be produced for substantially less compute than US labs had been spending, and that rigorous open-source release was a viable strategic position.