The Drake Passage is an ocean area between South America and Antarctic with very extreme wind and wave climates. The quality of forecasts of surface winds and significant wave heights is analyzed in the present paper, in order to study the error and distribution of model uncertainties as a function of: forecast range, severity, location, and numerical model. The operational metocean forecasts selected are run twice a day by the Brazilian Navy and the forecast horizon analyzed is five days. A period of one year (2017) is considered for the assessment. Observations consist of four satellite missions: JASON2, JASON3, CRYOSAT, and SARAL. The atmospheric models with 10-m winds are GFS (Global Forecast System) and ICON (Icosahedral Nonhydrostatic Model). They are used as inputs to force the wave model WAVEWATCH III with a mosaic of two grids. Results show that under calm to moderate conditions, within the first two days of forecast, the wind and wave model skill is very high. However, above the 90th percentile and beyond the third forecast day, the predictability drops significantly. It highlights specific contours of forecast range versus percentiles where the wind and wave modelers should focus, in order to anticipate and to improve the predictability of extreme events at the Drake Passage.