China launched an artificial intelligence-powered large language model on Tuesday, developed specifically for meteorological applications to offer smarter, more personalized and scenario-based weather services for the public and industries.

The Fenghe LLM, developed by the China Meteorological Administration's Public Meteorological Service Center, is designed to enhance weather services with a deeper understanding of meteorological services and access to authoritative data, said Zhu Xiaoxiang, director of the center.

"Its overall capabilities are comparable to those of a meteorological service professional with five years of experience," Zhu said.

By inputting weather-related questions into Fenghe's miniprogram or website, users can receive immediate AI-generated responses. These range from science popularization and weather queries to tailored suggestions for disaster alerts, transportation, tourism, logistics, energy and health.

For instance, in a self-driving travel scenario, the system acts as a travel adviser, automatically providing suggestions such as avoiding flooded highways during thunderstorms or recommending indoor destinations when severe weather is forecast.

According to Zhu, Fenghe adopted a foundation model that integrates general LLM capabilities with professional meteorological knowledge. It was trained on 550 billion tokens of meteorological data to ensure accurate understanding.

The system includes an intelligent agent development platform that connects real-time meteorological data systems such as Fenglei for nowcasting and Fengqing for medium — and short-range forecasting.

Li Juanzi, a professor at the Department of Computer Science and Technology at Tsinghua University, said Fenghe's design allows for deep integration with real-world applications.

"Our goal is to let AI meteorological services meet the needs of public life, administrative management and emergency coordination," she said.

Wang Jianghai, deputy director of the flood and drought disaster prevention department of the Ministry of Emergency Management, said the new AI system elevates disaster prevention and response capabilities.

"Fenghe is like an expert adviser for flood control workers, which is capable of quantifying weather patterns, analyzing disaster risks, and providing precise suggestions for efficient prevention and mitigation," Wang said.

"For the public, it serves as a reliable weather companion that can offer timely travel and safety tips for them to plan and adjust their trips, greatly enhancing public awareness," he added.

Fenghe will undergo continuous evaluation and optimization through a standardized testing system to ensure its long-term effectiveness, according to CMA.