Abstract A 30-year wave hindcast for the North Sea covering the period of 1990–2020 is described. The simulations are performed by nesting a 5 km SWAN wave model to a global WAVEWATCH III implementation. For the coastal nest several options of the ST6 physics are tested, including choices for the coefficients of whitecapping dissipation, wind drag formulation and scaling of wind speed U10 to friction velocity u*. The results are shown for the two most widely used wind reanalyses: the CFSR and ERA5. The model dataset is validated and calibrated against satellite and several buoy measurements, allowing for the assessment of model performance in the different areas of the North Sea. The ERA5 reanalysis outperforms the CFSR winds in all validation metrics except for a negative bias in wind speed, which is more noticeable for the upper quantiles. Similar results are observed for significant wave height when comparing model results forced by the two reanalysis winds. The ST6 source term performs considerably better than the source terms of van der Westhuysen et al. However, the results are sensitive to the choice of the ST6 tuning parameters, which are in turn also dependent on the forcing winds. Further tests are performed by correcting the wind stress to account for the observed wind speed bias, which is a part of the ST6 implementation. Finally, the long-term statistics of wave height are briefly presented, with emphasis on the wave height extremes.