Envision’s Tianji Weather Foundation Model Wins ECMWF AI Weather Quest, Sets New Global Benchmark

Envision, a global leader in green technologies, announced that its independently developed Weather Foundation Model, Tianji, secured first place overall in the latest edition of the AI Weather Quest, a competition organized by the European Centre for Medium‑Range Weather Forecasts (ECMWF). Tianji outperformed nearly 20 top teams worldwide and topped every evaluated category, including near‑surface variable mean temperature, mean sea‑level pressure, and precipitation for the 19‑ to 25‑day forecast window.

The AI Weather Quest evaluates models through continuous real‑time forecasting and verification, assessing sustainable performance across diverse weather conditions. Beyond the 15‑day medium‑range horizon, the competition focuses on probabilistic forecasts of weekly mean temperature, mean sea‑level pressure, and weekly precipitation for days 19–25. Temperature and precipitation are evaluated over predominantly land‑covered grid cells, while mean sea‑level pressure is assessed globally, emphasizing the ability to predict weekly conditions and their uncertainty weeks in advance. Teams submit weekly forecasts that are matched against actual data and publicly updated in the leaderboard.

This victory follows a series of top performances by Tianji, including a top spot in the global WeatherBench‑2 assessment, a tie with Google’s WeatherNext‑2 model, and first place overall in deterministic forecasts in the AI Weather Forecast Model Demonstration Program organized by China’s National Meteorological Center. Tianji also earned first place in State Grid’s AI Load Forecasting Competition and China Southern Power Grid’s business competition.

"Subseasonal forecasts – two weeks to two months ahead – are among the greatest challenges in meteorology," said Robert Zheng, head of Tianji at Envision. "The longer the horizon, the greater the uncertainty in atmospheric changes and the higher the demand for a model’s ability to detect evolving weather trends. Building on Envision’s years of experience in new energy, we developed the world’s first physics‑based AI weather model that uses petabyte‑scale high‑resolution spatial and temporal data and a three‑dimensional network from space, air, and ground to deliver reliable predictions across the entire period."

Powered by advanced computing, Tianji delivers fast short‑term and global forecasts with markedly improved accuracy for wind speed and solar irradiance in key regions. It covers the full spectrum from minute‑level predictions to subseasonal trend analysis, providing comprehensive meteorological information for real‑time monitoring and long‑term planning.

The win underscores Tianji’s capability to span the spectrum from short‑term forecasts to subseasonal predictions, setting a new global reference for physics‑based AI weather forecasting.