01 / AI · Analysis
Darwin claims lowest forecast error across 28 categories
Darwin’s performance claim rests on company-published figures comparing its AI weather engine with established forecast models.

A dev.to article reported that Darwin achieved the lowest error across all 28 categories in a company-published comparison with forecast models operated by national weather agencies and other forecasting bodies.
The comparison concerns errors in forecasts from an AI weather engine and established models. The models selected and the scope of the evaluation frame what the performance claim means.
According to the article, VIDRAFT, the developer of Darwin-Prediction-Weather, conducted the comparison. The results it reported are therefore company-published performance figures.
Weather agency models in the comparison
The dev.to article identified forecast models from the European Centre for Medium-Range Weather Forecasts and the US National Oceanic and Atmospheric Administration as comparison targets. Models from weather agencies in Japan, Canada, Germany and Norway were also included.
According to the same article, the comparison covered 24 cities. The company-published claim of the lowest error applies to the results for those cities.
What the category rankings show
The dev.to headline presented Darwin as having the lowest error in every comparison category, based on company-published figures. An error ranking and the size of the error gap between models convey different information.
The article identified the model’s developer as the organisation conducting the comparison. Assessing the performance claim therefore requires reading the rankings alongside the comparison conditions and the size of the errors.