Lineage · Since 2023

The DeepSeek lineage

DeepSeek's open-weight reasoning and coding model line.

Generation 1 · Dec 26, 2024

DeepSeek V3

671B MoE open-weight base model.

Open128K ctxMoEopen weights
Generation 2 · Jan 20, 2025

DeepSeek R1

Open-weight reasoning model that popularized reinforcement-learned chain-of-thought at scale.

Open128K ctxreasoningopen weights