Abstract

We introduce our first-generation reasoning models, DeepSeek-R1-Zero andDeepSeek-R1. DeepSeek-R1-Zero, a model trained via large-scale reinforcementlearning (RL) without supervised fine-tuning (SFT) as a preliminary step,demonstrates remarkable reasoning capabilities. Through RL, DeepSeek-R1-Zeronaturally emerges with numerous powerful and intriguing reasoning behaviors.However, it encounters challenges such as poor readability, and languagemixing. To address these issues and further enhance reasoning performance, weintroduce DeepSeek-R1, which incorporates multi-stage training and cold-startdata before RL. DeepSeek-R1 achieves performance comparable to OpenAI-o1-1217on reasoning tasks. To support the research community, we open-sourceDeepSeek-R1-Zero, DeepSeek-R1, and six dense models (1.5B, 7B, 8B, 14B, 32B,70B) distilled from DeepSeek-R1 based on Qwen and Llama.