The capability of CMIP5 models in representing the structure and regional impact of the North Atlantic Oscillation (NAO) and Pacific-North American (PNA) during boreal winter is tested.  For each model, low frequency mode patterns that are most NAO-like and PNA-like are identified and then separately categorized into clusters according to the relative correspondence among their spatial patterns (along with the observed pattern).  For both NAO and PNA, about 20% of the model patterns lie within the same cluster as the observed pattern.  Differences among clusters mainly consist of either spatial shifts or regional amplitude differences in the primary anomaly features.  While all models qualitatively replicate the basic structure of the PNA pattern, a small minority of models fails to replicate the NAO pattern. Interestingly, models with a well-resolved stratosphere generally perform more poorly than those without.  Existing model biases in representing the NAO and PNA patterns are demonstrated to have important consequences for representing associated regional variability in surface air temperature and storm track behavior.  Finally, it is found that model deficiencies in the NAO and PNA patterns are related to underlying inadequacies in replicating the climatological stationary waves.