DeepSeek

Profile Overview

DeepSeek (深度求索) is a Chinese artificial intelligence company founded in 2023 by Liang Wenfeng, a quantitative finance entrepreneur who previously co-founded High-Flyer Quant (幻方量化). Headquartered in Hangzhou, China, DeepSeek rapidly emerged as one of the world's most disruptive AI labs by demonstrating that frontier-level models can be built at a fraction of the cost of US competitors. The company gained global attention in December 2024 with the release of DeepSeek V3, a 671B-parameter Mixture-of-Experts (MoE) model that matched GPT-4-class performance while being trained for only ~$5.6 million — a cost roughly 10–20× lower than comparable models. In January 2025, DeepSeek released R1, a reasoning model employing large-scale reinforcement learning that achieved performance comparable to OpenAI's o1 on math, coding, and scientific reasoning benchmarks. DeepSeek's breakthrough efficiency, achieved through novel MoE architectures, multi-token prediction, and FP8 mixed-precision training, sent shockwaves through global markets in early 2025, challenging the prevailing narrative that frontier AI requires billions in compute investment. Rather than raising external venture capital, DeepSeek has been entirely funded by High-Flyer Quant's trading profits, with High-Flyer having invested an estimated $1.5 billion in AI compute infrastructure since 2019. DeepSeek operates under a permissive MIT license for its models and API services, making frontier AI capabilities accessible to developers worldwide at significantly lower cost than any comparable provider.

Flagship Offerings

Vetted AI Models