China Agricultural University on Monday launched the Shennong Large Model 3.0, described as the country's most comprehensive and data-rich artificial intelligence model for agriculture, marking a significant step toward the intelligent transformation of farming and research worldwide.

The model, together with the newly introduced Shennong AI Agent Platform, was unveiled at the opening ceremony of the 2025 World AgriFood Innovation Conference in Beijing, signaling a new stage in agricultural AI that is both efficient and accessible.

Sun Qixin, president of China Agricultural University, said the third-generation Shennong model is China's first large-scale AI system trained with the country's largest agricultural dataset to date, covering 90 percent of agricultural disciplines and 80 percent of agricultural scenarios.

Named after Shennong, a legendary figure in ancient China known for teaching farming, the model was developed by a team led by Wang Yaojun, an associate professor at the university's College of Information and Electrical Engineering. The Shennong Large Model series has evolved rapidly since its first version in 2023.

Specifically designed for agricultural applications, the model covers areas that include crop breeding, planting and livestock production. "Compared with general-purpose models, Shennong is far more specialized in answering agricultural questions," Wang said, adding that it ranks among the world's leading models in the agricultural sector.

The first version focused on agricultural question-and-answer functions, while version 2.0 introduced multimodal capabilities and real-world applications. "The new 3.0 version features 36 intelligent agents, emphasizing small size, high intelligence and low cost to make agricultural AI more practical and widely available for farmers and production entities," Wang said.

"Through a mobile app or mini program, farmers can access AI-powered guidance for crop management and pest control within seconds," he added.

The pest and disease identification agent can recognize 70 categories and more than 600 types of crop pests and diseases, offering instant advise comparable to consulting multiple plant protection experts.

The model is already being tested in Beijing, the Inner Mongolia autonomous region, and Shanxi and Heilongjiang provinces, with more than 100,000 active users. In Heilongjiang, for example, farmers can quickly identify corn leaf spot and receive immediate recommendations on fertilizers, pesticides and treatment methods, Wang said.

He added that the model can also accelerate scientific discovery for agricultural researchers and assist in data-driven breeding decisions. Among the 36 intelligent agents, the Wheat Breeding AI Assistant integrates national-level germplasm data and expert knowledge to support precision breeding, helping move the process from traditional trial-and-error to data-based innovation.

The 2025 World AgriFood Innovation Conference runs from Friday to Tuesday in Beijing's Pinggu district, drawing nearly 800 participants from dozens of countries to discuss resilient food supply and agrifood system transformation.

At the opening ceremony, several other achievements were also unveiled, including the inauguration of the China Agricultural University–Dong-E-E-Jiao Joint Industry Innovation Institute and the release of the Consultative Group on International Agricultural Research 2025 Innovation Report.