DeepSeek
DeepSeek is a Chinese AI lab that develops open-source large language and reasoning models, distributed via API and open weights, targeting developers and enterprises globally.
Evidence notes
Who uses it: Developers and enterprises integrating LLM capabilities via API or self-hosted open weights; secondary audience is Chinese state-aligned enterprises and cloud providers (Alibaba, Tencent) seeking domestic AI infrastructure.[23]+4Reported
Public signals: DeepSeek Eyes $7.4B Maiden Funding Round as Tencent and CATL Weigh Major Stakes[12]Reported
Product: Mixture-of-Experts (MoE) architecture with selective activation (49B active of 1.6T total parameters), KV cache compression, DeepSeek Sparse Attention (DSA) for long-context efficiency, and reinforcement learning for reasoning.
Source ledger
- [1]DeepSeek V4 Pro Deep Dive June 2026 | Apache 2.0, 1.6T MoE Pricing
- [2]DeepSeek CEO, Founder, Key Executive Team, Board of Directors & Employees
- [3]DeepSeek slated to raise $7 billion in maiden funding round, sources say | MarketScreener
- [4]DeepSeek's $45bn valuation is also Beijing's strategic statement
- [5]DeepSeek Is Raising $7 Billion to Make Open Source Extremely Capital Intensive