Competition Structure
The AI Weather Quest kicked off in March 2025. It is a long-term benchmarking and learning framework for AI/ML-based sub-seasonal forecasting divided in 13-week competitive forecasting periods. Participants can take part in any forecasting period, allowing new teams to join over time and existing teams to continue refining their models between periods.
AI Weather Quest current framework: from September 2026 onwards
From September 2026 onwards, the AI Weather Quest framework introduces two new forecast diagnostics.
During the SON 2026 bridge period, participants continue submitting forecasts for global variables (weekly-mean near-surface temperature, precipitation, mean sea level pressure), while also having the opportunity to test submissions for two new variables (Madden-Julian Oscillation phase probabilities and tropical storm days). These test submissions are intended to help teams familiarise themselves with the new variables and are not included in the public evaluation during the bridge period.
From DJF 2026 onwards, the two new variables (Madden-Julian Oscillation phase probabilities and tropical storm days) evaluation will appear in the public leaderboards.
See below the dates of the upcoming 2026-2027 competitive periods, based on forecast initialisation:
Participants can compete in as many 13-week competitive periods as they wish, maximising flexibility and opportunities for success. Between periods, participants can continuously improve their forecast models, enabling real-time iterative development.
Join the QuestPrevious framework: March 2025 to September 2026
The initial AI Weather Quest framework covered the launch and first forecasting year of the competition. Participants submitted forecasts for the following global variables: weekly-mean near-surface temperature, precipitation, mean sea level pressure.
It included an Initial Training Phase, designed to help participants become familiar with the tools, data and submission process, followed by four 13-week competitive forecasting periods: SON 2025, DJF 2025, MAM 2026 and JJA 2026.
All outputs from the initial framework, including participating teams, leaderboards and webinar recordings, remain available in the Results section of this website.
