DeepSeek is a Chinese AI lab, formally Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co., Ltd., that develops the DeepSeek model lines. Its site lists general-purpose V-series models, reasoning-focused R-series models, and specialized models for code, math, and vision, each linked to a public code repository.
The lab’s DeepSeek-R1 paper (submitted January 2025, later published in Nature) drew wide attention for showing that strong reasoning behavior can emerge from reinforcement learning without large human-annotated reasoning datasets. DeepSeek became internationally prominent for releasing competitive models openly and at notably lower reported cost than many Western counterparts.
Why business readers should care: DeepSeek upended assumptions that frontier AI requires the largest budgets and US-based labs, and its open releases pressured pricing and strategy across the entire industry.