On Nov 10, the " FertiSmart Model", led by the Information Center of the Ministry of Agriculture and Rural Affairs (MARA), was officially launched in Jiujiang, Jiangxi province, marking a new breakthrough in key technologies for smart agriculture. The model was co-developed by the Institute of Agricultural Resources and Regional Planning of the Chinese Academy of Agricultural Sciences (IARRP-CAAS), the Hefei Institutes of Physical Science of the Chinese Academy of Sciences (CAS), and MARA's Information Center.

The FertiSmart Model is built upon the "Nutrient Expert professional fertilization recommendation model" , which was developed over many years by Prof. He Ping's team at IARRP-CAAS. It integrates high-throughput soil intelligent detection robotic systems developed by Chief Engineer Wang Rujing's team at the CAS Hefei Institutes of Physical Science. Supported by a nationwide dataset of more than one million nutrient measurements for 28 major crops and enhanced by the DeepSeek large language model's knowledge distillation and reasoning capabilities, the system establishes a collaborative architecture between a "professional fertilization model" and a "cloud-based intelligent brain", forming an advanced multi-agent intelligent fertilization system. The development and application of this model aim to accelerate the implementation of science-based fertilization and inject strong momentum into the rapid advancement of smart agriculture in China.

The model offers significant advantages, including simplified recommendation processes, dynamic fertilization decision-making, and markedly improved service efficiency for farmers. Users only need to input crop type, target yield, and fertilizer type. Leveraging the foundational Nutrient Expert model, the system further integrates multi-modal data—such as soil, remote sensing and meteorological information—and employs DeepSeek's data mining, knowledge retrieval, generative capabilities and logical reasoning to provide a one-click customized fertilization plan. Each plan includes recommended fertilizer dosage, application timing and placement, as well as soil diagnostics. The model effectively bridges the data and resource gaps between smallholder farmers and large-scale operations, enabling customized strategies for every field. Recommendation efficiency and accuracy have been enhanced by 20-fold and 10-fold, respectively.

For diverse agronomic scenarios—including organic fertilizer substitution, straw incorporation, deep fertilization machinery, and water-fertilizer integration—the model incorporates differentiated parameters and operational plans based on specialized knowledge bases and large-model distillation, ensuring scientific validity and applicability across practices.

The R&D team has already established complete smart fertilization technology systems in locations such as the Jiansanjiang Agricultural High-tech Zone in Heilongjiang province and Changfeng county in Anhui province. Demonstration results show that, while reducing chemical fertilizer use by 18%–22%, yields increased by 6%–14% and farmers' income rose by 8%–12%. The technology is highly scalable and can be rapidly applied across ecological zones and farm types nationwide.