Meteorological AI Science Model "Fengyuan" Of CMA Released

The China Meteorological Administration (CMA) marked a significant advancement in atmospheric science on December 19, 2025, with the official release of “Fengyuan” (风元) V1.0. This launch represents a pivotal milestone in the nation’s meteorological history, signaling the establishment of an independently developed, end-to-end scientific model base. Crucially, the Fengyuan model is supported by full proprietary intellectual property rights, ensuring that China maintains technological sovereignty over its primary meteorological AI infrastructure (https://vertexaisearch.cloud.google.com/grounding-api-redirect/).

Advanced End-to-End Architecture

Fengyuan V1.0 distinguishes itself through an innovative end-to-end architectural design. Unlike traditional meteorological systems that often rely on fragmented processing stages, this model is capable of generating comprehensive global weather forecasts directly from raw observational data. By removing traditional intermediate steps, the architecture minimizes the potential for data loss that can occur during multi-stage processing. Furthermore, this streamlined approach significantly enhances the speed at which raw sensor data is translated into actionable, high-fidelity forecasts (https://vertexaisearch.cloud.google.com/grounding-api-redirect/

Integrated Modules for Holistic Learning

The core of the Fengyuan V1.0 model consists of two primary integrated modules: data assimilation and forecasting. In many legacy systems, data assimilation—the process of incorporating real-world observations into a model—is treated as a separate pre-processing task. Fengyuan, however, integrates these functions, allowing the AI to learn the complex relationships between current observations and future atmospheric states more holistically. This integration ensures that the transition from current data to future predictions is more fluid and representative of real-world atmospheric dynamics (https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGfjtxDm7M6ef5SDsrN5H-togfuhRNVK_iw_vv-3i0rlcfsGu2Fr67O_UVOW1JpMSRfXBOMil_2wpehChpX-6OXiI4zB1nQDU8H2wW145QtHqMytvdfDPb0eXHTuWe4N7mJRan6AhqV9IF9ghKq4ezcJtIwJ-vtcIknt70dKRrZ-2lzvgUk_g==).

Multi-Institutional Collaboration and Development

The creation of Fengyuan was the result of an intensive multi-institutional collaboration led by the Chinese Academy of Meteorological Sciences (CAMS) and the Xiong’an Meteorological AI Innovation Research Institute. The development process also leveraged the specialized expertise of the Shanghai AI Laboratory and Fudan University. This cross-disciplinary partnership allowed for the successful combination of traditional atmospheric science with cutting-edge advancements in deep learning and computer vision (https://vertexaisearch.cloud.google.com/grounding-api-redirect/

To support the continued evolution of this technology, the Earth System AI Forecasting Joint Laboratory was inaugurated in Xiong’an simultaneously with the model’s release. This new facility is designated as the primary hub for the ongoing training, refinement, and iteration of Fengyuan and its derivative models, ensuring a centralized location for sustained research and development (https://vertexaisearch.cloud.google.com/grounding-api-redirect/

The “Feng-” Family: A Comprehensive AI Ecosystem

The release of Fengyuan V1.0 was accompanied by significant upgrades to three other operational models within the CMA’s AI forecasting suite. Together, these models form the “Feng-” family, a comprehensive ecosystem designed to cover a wide range of meteorological timeframes and needs:

Fengqing: A global operational model designed for broad-scale weather patterns.
Fenglei: Focused on nowcasting, providing critical data on a minute-by-minute basis.
Fengshun: A climate-focused model tailored for seasonal forecasting and long-term trends.

This tiered ecosystem allows the CMA to provide specialized AI-driven services ranging from immediate short-term alerts to long-range climate projections (https://vertexaisearch.cloud.google.com/grounding-api-redirect/

Economic Applications and Energy Security

Beyond traditional weather reporting, Fengyuan is specifically designed to support emerging sectors of the modern economy. Its high-precision, localized data is essential for the burgeoning low-altitude economy, where it facilitates safer and more reliable drone operations. Additionally, the model plays a vital role in national energy security by providing the granular data necessary for the efficient management of renewable energy grids, such as wind and solar power (https://vertexaisearch.cloud.google.com/grounding-api-redirect/

The model also extends its utility to health meteorology, offering insights into how weather patterns may impact public health. By providing more accurate forecasts, Fengyuan assists in mitigating weather-related risks across various industrial and social sectors (https://vertexaisearch.cloud.google.com/grounding-api-redirect/

International Leadership and Open Science

China intends to use Fengyuan as a vehicle for international cooperation and leadership in climate services. The CMA has announced that the model will be utilized to provide advanced meteorological services to countries participating in the Belt and Road Initiative (BRI). This initiative positions China as a global leader in providing AI-driven disaster prevention and climate resilience capabilities to developing nations (https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGYyukC_qCWveZ0vmtzXMlbcZ4J1L-rDAB74__jPmrQBbNYC91rHk_XwibdHhUkwyCgRGpIGTEasQAZBjKVlGyW4dJtTmvTonl5-51dsWP_tDkhAZ0XCCIcjTxbhFqbfHFMzWpbDTD_xrNA_w5hqm99zz0_iSqDGmvXCwc=).

Furthermore, Fengyuan is intended to serve as an open-source platform. By making the model base available to the global scientific community, the CMA aims to attract international industry experts and consolidate research and development efforts. This open-source strategy is expected to facilitate faster iteration of forecasting algorithms through community feedback and collaborative innovation (https://vertexaisearch.cloud.google.com/grounding-api-redirect/

Future Evolution: Integrating Physical Mechanisms

While the current iteration of Fengyuan is a powerful data-driven tool, the CMA has outlined a long-term goal for the model to evolve into a multi-sphere coupled system. This future version will move beyond pure AI patterns to integrate fundamental physical mechanisms, such as the conservation of mass and energy. By embedding these physical constraints directly into the AI’s learning process, the CMA ensures that future forecasts remain meteorologically consistent and physically plausible, combining the speed of AI with the rigor of classical physics (https://vertexaisearch.cloud.google.com/grounding-api-redirect/

Conclusion

The release of the Fengyuan V1.0 meteorological AI science model represents a transformative shift in how China approaches weather forecasting and climate science. Through its end-to-end architecture, integrated data modules, and multi-institutional foundation, the model provides a robust platform for both domestic economic support and international collaboration. As it moves toward a more physics-integrated future, Fengyuan is set to remain at the forefront of the global effort to harness artificial intelligence for the benefit of atmospheric science and disaster prevention.