OneRec Technical Report

Recommender systems have been widely used in various large-scale user-oriented platforms for many years. However, compared to the rapid developments in the AI community, recommendation systems have not achieved a breakthrough in recent years. For instance, they still rely on a multi-stage cascaded architecture rather than an end-to-end approach, leading to computational fragmentation and optimization inconsistencies, and hindering the effective application of key breakthrough technologies from the AI community in recommendation scenarios.To address these issues, we propose OneRec, which reshapes the recommendation system through an end-to-end generative approach and achieves promising results. Firstly, we have enhanced the computational FLOPs of the current recommendation model by 10 $\times$ and have identified the scaling laws for recommendations within certain boundaries. Secondly, reinforcement learning techniques, previously difficult to apply for optimizing recommendations, show significant potential in this framework. Lastly, through infrastructure optimizations, we have achieved 23.7% and 28.8% Model FLOPs Utilization (MFU) on flagship GPUs during training and inference, respectively, aligning closely with the LLM community. This architecture significantly reduces communication and storage overhead, resulting in operating expense that is only 10.6% of traditional recommendation pipelines. Deployed in Kuaishou/Kuaishou Lite APP, it handles 25% of total queries per second, enhancing overall App Stay Time by 0.54% and 1.24%, respectively. Additionally, we have observed significant increases in metrics such as 7-day Lifetime, which is a crucial indicator of recommendation experience. We also provide practical lessons and insights derived from developing, optimizing, and maintaining a production-scale recommendation system with significant real-world impact.

@article{zhou2025_2506.13695,
  title={ OneRec Technical Report },
  author={ Guorui Zhou and Jiaxin Deng and Jinghao Zhang and Kuo Cai and Lejian Ren and Qiang Luo and Qianqian Wang and Qigen Hu and Rui Huang and Shiyao Wang and Weifeng Ding and Wuchao Li and Xinchen Luo and Xingmei Wang and Zexuan Cheng and Zixing Zhang and Bin Zhang and Boxuan Wang and Chaoyi Ma and Chengru Song and Chenhui Wang and Di Wang and Dongxue Meng and Fan Yang and Fangyu Zhang and Feng Jiang and Fuxing Zhang and Gang Wang and Guowang Zhang and Han Li and Hengrui Hu and Hezheng Lin and Hongtao Cheng and Hongyang Cao and Huanjie Wang and Jiaming Huang and Jiapeng Chen and Jiaqiang Liu and Jinghui Jia and Kun Gai and Lantao Hu and Liang Zeng and Liao Yu and Qiang Wang and Qidong Zhou and Shengzhe Wang and Shihui He and Shuang Yang and Shujie Yang and Sui Huang and Tao Wu and Tiantian He and Tingting Gao and Wei Yuan and Xiao Liang and Xiaoxiao Xu and Xugang Liu and Yan Wang and Yi Wang and Yiwu Liu and Yue Song and Yufei Zhang and Yunfan Wu and Yunfeng Zhao and Zhanyu Liu },
  journal={arXiv preprint arXiv:2506.13695},
  year={ 2025 }
}