China unveils AI model to revolutionize soybean breeding

According to Mettis Global, Chinese researchers have introduced Fengshu 2.0 at a national soybean research symposium in Hefei, Anhui Province, citing Anhui Agricultural University. The artificial intelligence model was built for soybean research and breeding by that university together with the Institute of Crop Sciences of the Chinese Academy of Agricultural Sciences. It draws on a wide range of material, including germplasm resources, genomic and coding sequences, proteins, gene expression data, disease samples, breeding trial records and both domestic and foreign research literature.

The platform links several general-purpose language models, and a dedicated module compares and merges their answers using domain knowledge and structured evaluation rules. Fengshu 2.0 can help with disease diagnosis, parent selection, virtual cross design, marker-assisted breeding and gene function analysis. The first version appeared in February, and the new release aims to move breeding away from experience-led and trial-led methods toward data-backed decisions.

Developers intend to add climate, soil and cropping-pattern data from China's major soybean-producing regions, while widening field trials. The platform is expected to improve germplasm discovery, trait assessment and variety selection, supporting soybean varieties with higher yields, better quality and greater stress tolerance.