In the office of Wang Yaojun, an associate professor at China Agricultural University, a white device about the size of a microwave draws immediate attention. The machine functions as a miniature smart farm powered by the Shennong Large Model, an artificial intelligence system developed by the university to support agricultural production.
At the center of the device is a proprietary AI model equipped with 36 specialized intelligent agents. The system currently serves more than 100,000 farmers across China, offering decision support across planting, cultivation, and crop protection.
To build the model’s core knowledge base, Wang and his team relied on the university library and public agricultural resources. Over seven months, faculty members and students digitized more than 3,000 agricultural books. Combined with public datasets, the effort produced a specialized database of 20,000 volumes tailored to farming applications.
The team also conducted field research across more than 20 provinces. They collected data on soil composition, irrigation practices, pest and disease outbreaks, and extreme weather impacts, covering the full agricultural production cycle.
From Pilot Model to Field Adoption
The first version of the Shennong Large Model was released in December 2023, offering agricultural question-and-answer services, semantic analysis of farming texts, text summarization, and operational decision support. Version 3.0, launched in October 2025, added 36 scenario-specific intelligent agents designed to guide cultivation practices and reduce the technical barrier for users.
The system now includes a proprietary knowledge framework covering about 90 percent of agricultural disciplines and 80 percent of farming scenarios. Its core database contains 10 million knowledge graph entries, 20 million annotated images, and 50 million production records. The model is currently being piloted in several regions, effectively acting as an AI-based management system for modern agriculture.
Farmers report measurable results. Du Lianhui, who manages 3,000 mu of farmland in Liaoning Province, began using the Shennong model on 600 mu of corn in 2025. Sensors deployed across his fields transmit real-time data on temperature, sunlight, pests, and disease conditions to the platform. Du said production costs fell from about 480 yuan per mu to under 400 yuan within a year.
To address adoption challenges, particularly among older farmers, the research team developed a WeChat mini-program that simplifies access to the model’s core functions. Farmers can upload photos of crops to identify diseases and receive treatment recommendations within seconds.
Wang said the success of agricultural AI depends not only on technical capability but also on usability and trust. The Shennong project reflects a growing effort in China to apply AI systems directly to traditional industries with practical, field-tested tools.
