China Launches ‘Mazu’ AI Toolbox for Satellite Weather Forecasting
China’s Meteorological Administration (CMA) has officially released the “Mazu” Fengyun Satellite AI Toolbox, a new system that integrates Fengyun meteorological satellite observation capabilities with artificial intelligence to provide international users with an integrated satellite application solution. The announcement was made on July 17 at the 2026 World Artificial Intelligence Conference (WAIC) in Shanghai, marking a significant milestone in China’s AI-driven meteorological science.
A New Era for AI-Powered Weather Forecasting
The Mazu Fengyun Satellite AI Toolbox (MAZU-FengYun Satellite AI Box) combines satellite data reception, AI inference, multi-source data fusion, and operational applications into a single “one-stop” closed-loop system, according to People’s Daily. The toolbox can be rapidly deployed across various scenarios, supporting internet users, Fengyun satellite direct receiving station users, CMA’s satellite broadcast system (CMACast), and field emergency operations.
A key innovation is the use of edge computing, which deploys AI models directly to application terminals. This enables local data processing and intelligent analysis, addressing a long-standing challenge where countries could access satellite data but lacked the network infrastructure, computing power, or technical expertise to convert it into actionable weather services. As Xinhua reported, the toolbox helps users quickly develop locally-tailored meteorological service products, bringing advanced weather technology to the grassroots level.
From 7 Countries to 30: Expanding Global Reach
The launch of the AI toolbox coincides with a major expansion of the broader MAZU program. During the WAIC opening ceremony, President Xi Jinping announced that China would promote the MAZU intelligent early warning solution for deployment in 30 countries, up from the seven nations where it is already operational — Pakistan, Ethiopia, Solomon Islands, Jordan, Sri Lanka, Mongolia, and Djibouti, with cloud trials underway in more than 40 additional countries.
“MAZU represents a seamless integration of our excellent traditional culture with modern technology,” CMA head Chen Zhenlin said in a previous interview with Qiushi Journal and China Daily. “It is designed to provide customized warning systems and proven disaster prevention wisdom to developing nations.”
Djibouti 2.0 and Thailand Collaboration
At the same event, the CMA handed over the upgraded “MAZU Djibouti 2.0” solution to Djibouti. The new version improves forecast resolution from 9 kilometers to 3 kilometers, provides forecasts up to three days in advance, and updates them every six hours. As Xinhua English reported, the system uses phased-array radar, AI forecasting models, and Fengyun satellites to improve extreme weather monitoring and early warning.
“Some areas of Djibouti have long faced threats from natural disasters such as floods, heat waves, and cyclones,” said Mohamed Ismail Nour, Director of the Djibouti National Meteorological Agency, as quoted by The Paper. “After learning about the latest progress of Djibouti 2.0, we believe this model will greatly enhance our ability to respond to frequent disasters.”
Additionally, the CMA announced the launch of the China-Thailand Joint Laboratory for Meteorological Disaster Intelligent Prediction and Warning — the world’s first bilateral international cooperation laboratory in meteorological AI.
Fenghe LLM and the AI Model Ecosystem
The CMA also released “Fenghe,” an AI meteorological service large language model trained on 50 million tokens of high-quality meteorological service data, with a global open-source plan announced. The MAZU ecosystem now includes the “Fenglei,” “Fengqing,” and “Fengshun” AI forecasting models covering timescales from minutes to seasons.
International Recognition and Strategic Significance
UN Secretary-General António Guterres praised China’s MAZU program during the conference, stating that “China’s platforms, satellites, and AI models have helped dozens of countries improve their early warning capabilities.” Guterres described the program as “exactly the cooperation model the world needs — helping countries protect lives through technology transfer, joint research, and local capacity building.”
The MAZU program — whose name is both an acronym (Multi-hazard, Alert, Zero-gap, Universal) and a reference to the Chinese sea goddess Mazu, a mythological protector against maritime disasters — represents a comprehensive AI-powered multi-hazard early warning solution. It combines Fengyun satellite data, AI forecasting models, edge computing, and customizable warning systems.
What’s Next
With the MAZU Djibouti 2.0 expected to become operational by the end of 2026 and the expansion to 30 countries underway, the program is positioned as a flagship technology diplomacy initiative. The CMA has committed to training 2,000 people, providing 100 scholarships, and hosting 50 visiting scholars over the next three years. As the program scales, key questions remain regarding effectiveness metrics, long-term sustainability, and the governance framework for the open-source Fenghe LLM.