Dear all,
We are pleased to share the results and outputs from the DJF (December–February) evaluation period.
Results
MicroEnsemble (Microsoft · Harvard University · University of Toronto · IMPA · Massachusetts Institute of Technology · Rhiza Research) leads the rankings for both week-3 and week-4 forecasts.
Six additional teams also outperform climatology:
-
LP (Jiangsu Climate Center)
-
AIFS (ECMWF)
-
CMAandFDU (China Meteorological Administration · Fudan University)
-
UWAtmos (Chinese Academy of Sciences · NVIDIA · University of Washington)
-
CLiMA (Climate Modeling Alliance)
-
scienceAI (anonymous team)
See the full leaderboards: https://lnkd.in/ewVpaij7
DJF forecast data release
All DJF-period forecast data is now publicly available for independent verification, reproducibility, and exploratory analysis via the AI Weather Quest Data Portal. Team-specific datasets are also accessible from each team’s page.
Methodology summaries published
Teams have shared concise descriptions outlining model architectures, input data, and training or ensemble strategies, available on their respective team pages: Teams • AI Weather Quest.
First scientific insights
Read the initial analysis of the competition results in our science blog:
https://lnkd.in/eaVGgGsF
DJF Awards Webinar – 19 March 2026, 13:00 UTC
Hear directly from MicroEnsemble, LP and AIFS as they present their AI approaches for weeks 3–4 forecasting: https://lnkd.in/dUq6m7Cr
What’s next
The third forecasting period (March–May) is now underway, and new teams are welcome to join at any time.
Many thanks to all DJF participants for their continued contributions to advancing AI-based sub-seasonal prediction.