Frequently Asked Questions About DeepSeek Large Language Model (LLM)

Tenable’s Security Response Team provides an FAQ on DeepSeek LLMs, detailing what they are, differences between V3 and the reasoning-centric R1 (MoE, MLA, MTP/CoT), available distilled sizes and hardware needs for local deployment, and how they compare with other frontier models. It emphasizes privacy and security risks of using DeepSeek’s hosted services due to data collection and storage under PRC jurisdiction, notes bans or reviews by several countries and U.S. government entities, and observes that local, offline use reduces exposure while Tenable continues LLM safety research.