ORLANDO, Fla. – While the 2025 Atlantic hurricane season may not have been the busiest in recent years, it did come with some noteworthy storms.
The season brought three Category 5 hurricanes, the second-highest total on record, with the bulk of the activity coming after the traditional peak in mid-September. That late-season surge, combined with several rapid intensification events, made forecasting extra challenging.
Behind every one of those forecasts is a combination of science-driven data and human experience.
While no single computer model gets it right every time, forecasters use it as a guide for forecasting a hurricane’s track and intensity.
According to a new report from the NOAA National Hurricane Center, track forecasts were a strong point in 2025, continuing a decades-long trend of improvement.
Forecast errors in 2025
Errors were lower than the recent average at all timeframes.
Track forecasts were better than the recent 5-year average at every time frame, continuing a long-term trend of improvement.
But intensity forecasts were tougher, with errors running significantly higher than normal, largely due to frequent rapid strengthening.
Top-performing models in 2025
Forecast track:
- GDMI (Google DeepMind) stood out as the top performer
- The official forecast (OFCL) performed and was the most consistent from run to run.
- HCCA, HWFI & FSSE (hurricane/consensus models) were consistently strong and among the most reliable overall.
- Models like GFS and HWRF struggled more, especially at longer ranges.
Forecast intensity:
- GDMI (Google DeepMind) again led the pack, showing high accuracy across most timeframes.
- The official forecast (OFCL) remained a top player running just below AI modeling.
- HFAI and HFBI (hurricane-specific models) performed well, especially in the short to medium range.
- GFS came in practically in last place in its intensity forecasting, too.