In this study, the control simulations of two general circulation model (GCM) experiments are assessed in terms of their ability to reproduce realistic ‘real world’ weather. The models examined are the UK Meteorological Office high-resolution atmospheric model (UKHI) and a coupled ocean/atmosphere model of the Max Planck Institut für Meteorologic, Hamburg (MPI). An objective classification of daily airflow patterns over the British Isles is used as a basis for comparing the frequencies of model-generated weather types with the frequencies derived from 110 years of observed mean-sea-level pressure (MSLP) fields. The weather-type frequencies generated by the GCMs, and their relationships with simulated monthly mean temperatures and total precipitation over the UK, are compared, season by season, with similar results derived using the observational data. An index of gale frequencies over the British Isles, derived from a similar objective analysis of daily MSLP fields, is used to evaluate the ability of the GCMs to simulate the observed frequency of storm events. One advantage of using 110 years of observational data is that the observed decadal-scale variability of climate can be introduced into this type of validation exercise. Both the GCMs assessed here are too cyclonic in winter. The seasonality of both anticyclonic and cyclonic types is much too strong in MPI and summer precipitation in this model is greatly underestimated. MPI simulates the annual cycle of temperature well, while UKHI successfully reproduces the annual cycle of precipitation. The analysis also indicates that the summer temperature variability of the two models is not driven by circulation changes.