The DeepSeek lineage

Follow a single frontier lab's innovations in order: the math-RL algorithm they invented, the attention and MoE efficiency tricks, and the reasoning model those choices enabled.

• DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
• DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model
• Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention
• DeepSeek-V3
• DeepSeek-R1