Deepseekr1 Incentivizes Reasoning In Llms Through Reinforcement Learning Daya Guo Dejian Yang Haowei Zhang Junxiao Song Peiyi Wang Qihao Zhu Runxin Xu Ruoyu Zhang Shirong Ma Xiao Bi Xiaokang Zhang Xingkai Yu Yu Wu Z F Wu Zhibin Gou Zhihong Shao Zhuoshu Li Ziyi Gao Aixin Liu Bing Xue by Daya Guo & Dejian Yang & Haowei Zhang & Junxiao Song & Peiyi Wang & Qihao Zhu & Runxin Xu & Ruoyu Zhang & Shirong Ma & Xiao Bi & Xiaokang Zhang & Xingkai Yu & Yu Wu & Z. F. Wu & Zhibin Gou & Zhihong Shao & Zhuoshu Li & Ziyi Gao & Aixin Liu & Bing Xue &... instant download after payment.

General reasoning represents a long-standing and formidable challenge in artifcial Received: 14 February 2025intelligence (AI). Recent breakthroughs, exemplifed by large language models (LLMs)1,2 and chain-of-thought (CoT) prompting3, have achieved considerable Accepted: 17 July 2025success on foundational reasoning tasks. However, this success is heavily contingent Published online: 17 September 2025on extensive human-annotated demonstrations and the capabilities of models are Open accessstill insufcient for more complex problems. Here we show that the reasoning Check for updatesabilities of LLMs can be incentivized through pure reinforcement learning (RL), obviating the need for human-labelled reasoning trajectories. The proposed RL framework facilitates the emergent development of advanced reasoning patterns, such as self-refection, verifcation and dynamic strategy adaptation. Consequently, the trained model achieves superior performance on verifable tasks such as mathematics, coding competitions and STEM felds, surpassing its counterparts trained through conventional supervised learning on human demonstrations. Moreover, the emergent reasoning patterns exhibited by these large-scale models can be systematically used to guide and enhance the reasoning capabilities of smaller models.