Today, the frontier for weather prediction is in the subseasonal-to-seasonal (S2S) range - i.e., between 2 to 8 weeks lead time. A fundamental premise for S2S forecasting is that the atmospheric circulation can be characterized generally by a finite number of recurrent, quasi-stationary flow patterns, referred to as weather regimes. Past research has identified such weather regimes for the Atlantic-European sector and shown that state-of-the-art NWP systems provide skillful predictions of weather regimes out to two weeks, but not always. Recently, the PI and his research team derived unique winter weather regimes for North America. However, how well these weather regimes are predicted in S2S operational models along with if these models can replicate the associated sensible weather patterns characteristic of each regime remain unknown. This presentation quantifies the prediction skill of North American winter weather regimes in reforecasts from the S2S Project Database and examines process-oriented diagnostics to understand why the models behave as they do. Evidence indicates that the operational subseasonal models represent well the large-scale defining characteristics of each of the regimes, but the magnitude and duration of the regime lifecycles are less consistent with reanalysis. These limitations subsequently translate into shorter-than-observed sensible weather patterns and their overall impact on North American winter weather. Moreover, the models generally feature a prediction limit of 14-20 days lead time for all regimes, with the longest predictive window seen for the Alaskan Ridge regime. Investigations into “forecasts of opportunity,” whereby we link regime predictability to features like stratospheric polar vortex variability and warm conveyor belt activity in the Pacific associated with extratropical cyclones, are also performed. As such, we identify potential times when higher subseasonal forecast skill for North American winter weather may be possible and offer a process-oriented diagnosis of the models that may help future their development for skillful subseasonal forecasts.