DeepSeek

DeepSeek is a Hangzhou-based AI lab spun out of the quantitative hedge fund High-Flyer and led by founder Liang Wenfeng. Its January 2025 release of the R1 reasoning model — open-weight, MIT-licensed, and trained at a fraction of reported Western costs — triggered a global market selloff and forced a rethink of AI capital-efficiency assumptions. It remains the standard-bearer for low-cost, open-weight frontier research from China.

DeepSeek was founded in 2023 by Liang Wenfeng, whose hedge fund High-Flyer had stockpiled NVIDIA GPUs for quantitative trading research before US export controls tightened. Operating more like a research collective than a commercial vendor, the lab publishes detailed technical reports and releases model weights under permissive licenses.

Its V3 model demonstrated that a frontier-class mixture-of-experts model could be trained for single-digit millions of dollars in marginal compute cost, and R1 showed that reinforcement-learning-driven reasoning could be replicated openly. The "DeepSeek moment" of January 2025 briefly erased hundreds of billions of dollars in US chip and AI market capitalization and pushed every major lab toward more aggressive pricing.

DeepSeek's API is among the cheapest of any frontier-class provider, and its open weights are served by inference platforms worldwide — often at higher throughput than DeepSeek's own infrastructure, which is constrained by export controls on advanced accelerators. Alongside Alibaba's Qwen, it anchors China's open-weight ecosystem and its growing influence on global model development.