China Meteorological Administration launches global open-source plan for 'Fenghe' AI model

China's Meteorological Administration (CMA) has launched a global open-source initiative for its 'Fenghe' AI model, announced as the world's first weather-focused large language model with over one hundred billion parameters. The announcement was made during the World Artificial Intelligence Conference (WAIC) on July 17.

The 'Fenghe' model, developed in collaboration with entities like the Xiong'an AI Innovation Research Institute and ZhiPu, seeks to advance the use of AI in meteorological services and support global cooperation in disaster prevention and sustainable development. It is designed to provide intelligent support for weather analysis, risk assessment, and service delivery.

The model was trained on a dataset of over 50 million tokens of high-quality meteorological service data and authoritative information. CMA has made the complete model weights and standardized API interfaces publicly available on platforms such as GitHub and Hugging Face, allowing for integration into various applications, including mobile apps and mini-programs.

'Fenghe' currently supports national meteorological services and offers personalized weather information and risk warnings to the public in China. An international version, supporting both English and Chinese, is already integrated into the 'Mazu' early warning system, providing intelligent Q&A, weather queries, and risk analysis for global users.

Through this initiative, CMA aims to foster an open ecosystem for meteorological AI. The global community can leverage the model to develop new risk products, translate complex warnings into understandable formats, and integrate them with other sectors like transportation and energy, thereby strengthening early warning systems.