The Feiyu-1.0, a South China Sea ocean-atmosphere two-way coupled intelligent large model was officially released in Guangzhou on February 7. Jointly developed by the South China Sea Institute of Oceanology under the Chinese Academy of Sciences and China University of Petroleum (East China), it is the world's first intelligent large model specifically designed for two-way ocean-atmosphere coupling in the South China Sea.

Feiyu-1.0 delivers three major technological breakthroughs: autonomous and controllable core data, intelligent two-way ocean-atmosphere coupling, and low-cost, flexible, and scalable "plug-and-play" learning. The model is trained on a high-resolution South China Sea reanalysis dataset, independently developed by the South China Sea Institute of Oceanology. This marks the first time a domestic ocean AI model is no longer reliant on foreign core datasets, breaking long-standing dependence on European and U.S. reanalysis products, while enabling precise representation of small-scale processes such as internal waves and ocean fronts.

Most earlier AI models focused exclusively on either atmospheric or oceanic systems, treating the two as largely independent. Feiyu-1.0 integrates physical mechanisms with AI, adopting a multi-expert-system-based Swin Transformer architecture. Through a dual "fast-slow" learning framework, the model intelligently simulates two-way exchanges of momentum, heat, and other key variables between the ocean and atmosphere, significantly improving forecast accuracy for critical coupled elements.

Feiyu-1.0 is expected to play an expanding role in marine environmental protection, resource development, and climate change response, helping advance a smarter and more sustainable paradigm for ocean governance.