DEHB is a new Hyperparameter Optimization method that combines the advantages of Hyperband and Differential Evolution, and achieves strong performance far more robustly on a broad range of HPO problems and tabular benchmarks, especially for high-dimensional problems with discrete input dimensions.