China’s Nanjing Agricultural University (NAU) has launched Sinong, the country’s first open-source vertical large language model (LLM) designed specifically for the general agricultural sector, marking a major advance in the application of foundational AI to agriculture.

According to Xinhua News Agency, the release represents a significant breakthrough in China’s agricultural AI research, enabling more specialised and reliable use of large language models across the farming value chain.

Sinong has been trained on an extensive, highly structured agricultural dataset spanning multiple disciplines, including animal science, agricultural economics and management, agricultural resources and environmental science, horticulture, smart agriculture, veterinary medicine, plant protection, and crop breeding.

Named after ancient Chinese officials responsible for agriculture and finance, the model integrates knowledge from nearly 9,000 books, more than 240,000 academic papers, around 20,000 policy documents and technical standards, as well as large volumes of curated web-based content.

To address common challenges faced by specialised LLMs, such as hallucinations and outdated knowledge, the development team implemented a series of advanced technical measures. In addition to standard instruction fine-tuning, the training process incorporated multi-dimensional data inputs, including chain-of-thought reasoning and contextual references, significantly improving the model’s understanding and generation of professional agricultural content, Xinhua reported.

Sinong has been fully open-sourced and is now available on platforms including ModelScope and GitHub. NAU said the open-source approach is intended to lower barriers to AI adoption in agriculture, enabling research institutions, enterprises, and developers to carry out secondary development and drive innovation, ultimately supporting the growth of a collaborative ecosystem for smart agriculture solutions.