China Launches Global AI Model to Enhance Meteorological Services

China’s meteorological authority on July 17 unveiled Fenghe, described as a billion‑parameter, open‑source meteorological large language model designed to support weather analysis, risk assessment and public early‑warning services. The China Meteorological Administration said the model was developed with national earth‑system data resources and trained on a high‑quality meteorological corpus to provide multilingual intelligent query, risk analysis and decision support.

What Fenghe Does and How It Will Be Shared

Fenghe is intended to convert complex meteorological data into accessible, actionable outputs — including natural‑language risk summaries, localized early‑warning messages and technical support for professional forecasters. The CMA announced that complete model weights, standardized APIs, cloud services and deployment packages will be published on platforms such as GitHub, Hugging Face and ModelScope to enable global adoption and local customization.

International Cooperation and Early Warning Integration

China framed the launch as part of a broader push to strengthen global early‑warning capacity and disaster risk reduction. The model’s international edition has been integrated with China’s MAZU intelligent early‑warning framework and is being promoted for use in UN‑linked initiatives and bilateral cooperation, including a planned China‑Thailand joint laboratory for AI meteorological prediction. Officials said the effort targets improved forecasting for typhoons, heavy rainfall, heatwaves and drought at sub‑seasonal to seasonal timescales.

Technical and Operational Notes

The CMA reported that Fenghe was trained on a 50‑million‑token meteorological corpus and integrates authoritative observational and model datasets, and that the model has completed required filings for generative AI deployment to ensure safety and controllability. The open‑source release includes tools for embedding the model in apps, small programs and edge devices to broaden access.

Implications for Developing Countries and Humanitarian Response

Experts say an open, domain‑specific meteorological LLM could accelerate the conversion of raw data into locally relevant warnings, helping vulnerable countries improve preparedness and response. China highlighted recent deliveries of upgraded early‑warning systems to partner countries and said Fenghe will support capacity building and joint research on long‑range hazard prediction.