A new AI weather forecasting engine named Darwin-Prediction-Weather, developed by Korean startup VIDRAFT, has demonstrated superior performance against established meteorological models. In a nine-month study across 24 cities, Darwin-Prediction-Weather outperformed eight leading numerical and AI-based forecast models, including those from ECMWF and NOAA, across all 28 evaluated categories. Notably, for 7-day maximum temperature forecasts, Darwin achieved a 14% lower Root Mean Square Error than the next best model, ECMWF AIFS, and its effective forecast horizon for this metric was one day longer than ECMWF's AI model. AI
IMPACT Sets a new benchmark for AI in weather forecasting, potentially accelerating adoption in climate tech and geospatial ML.
RANK_REASON AI model achieves benchmark results surpassing established industry leaders. [lever_c_demoted from research: ic=1 ai=1.0]
- Darwin-Prediction-Weather
- Deutscher Wetterdienst
- ECMWF AIFS
- Environment and Climate Change Canada
- ECMWF
- Google GraphCast
- Google Research
- Huawei Pangu-Weather
- NOAA
- Norwegian Meteorological Institute
- VIDRAFT
- WeatherBench2
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →