The China Meteorological Administration (CMA) has put an AI-based global ensemble forecasting system into national operational use. The Fengqing artificial intelligence global short- and medium-range ensemble weather forecast system, designated CMA-AIM-GFS-Fengqing-ENS-V1.0, recently passed an operational review held at the National Meteorological Centre, according to Chinese-language reports from China Science Daily and IT Home.
The new system should not be confused with the original Fengqing model. That earlier version, developed by the National Meteorological Centre with Tsinghua University in 2024, is a deterministic model that produces a single forecast run. According to a paper in a CMA journal, it has about 3 billion parameters and generates 15-day global forecasts at 25-kilometer resolution in roughly three minutes. The new release is an ensemble system: it produces many forecast members to describe the uncertainty in atmospheric motion, which is the basis for probability forecasts of heavy rain, typhoons and severe convection.
The National Meteorological Centre led development together with the Xiong'an Meteorological Artificial Intelligence Innovation Institute, the National Meteorological Information Centre and Sugon. Reports say the model combines a vision Transformer architecture with physical constraints from meteorology, and that its data, algorithms and computing were all sourced domestically. Training and iteration ran on an AI-for-science cluster of about 100,000 accelerator cards at the National Supercomputing Center in Tianjin.
On performance, CMA cites independent verification on samples from January 2025 to August 2026. The effective forecast range for the ensemble mean of global 500 hPa geopotential height reached 11.0 days. The developers say key ensemble metrics, including the continuous ranked probability score of members and the spread-error ratio, performed well, and that some indicators beat both the European Centre for Medium-Range Weather Forecasts numerical ensemble and its AI ensemble system, AIFS-CRPS. These comparisons are the developers' own; detailed scores have not been published.
Like other AI forecast models, the system can complete a global run in minutes once trained. CMA says it will serve disaster prevention, transport, energy dispatch and ecological protection, and that the team will continue improving accuracy and lead time.
SEE ALSO: MAZU AI Weather System Goes Global · Xiandao Cup Student AI Computing Competition