High-latitude climate change is expected to increase the demand for reliable weather and environmental forecasts in polar regions. In this study, a quantitative assessment of the skill of state-of-the-art global weather prediction systems in polar regions is given using data from the THORPEX Interactive Grand Global Ensemble (TIGGE) for the period 2006/2007-2012/2013. Forecast skill in the Arctic is comparable to that found in the Northern Hemisphere midlatitudes. However, relative differences in the quality between different forecasting systems appear to be amplified in the Arctic. Furthermore, analysis uncertainty in the Arctic is more of an issue than it is in the midlatitudes, especially when it comes to near-surface parameters over snow- and ice-covered surfaces. Using NOAA's reforecast dataset, it is shown that the changes in forecast skill during the 7-year period considered here can largely be explained by flow-dependent error growth, especially for the more skilful forecasting systems. Finally, a direct comparison between the Arctic and Antarctic suggests that predictions of mid-topospheric flow in the former region are more skilful.