The evaluation of a climate model's capability to simulate trends in extreme events is not straightforward. This is due to the role of internal climate variability, the simulated phases of which are unique to each individual model realization. We undertake an assessment of the 21-member Community Earth System Model (CESM) on the basis of its ability to simulate heat wave days frequency over Australia. We employ the extreme heat factor definition to measure heat waves and study events for all summers (November-March) between 1950 and 2005. The spatial pattern, magnitude, and significance of trends in CESM were found to be reasonable since the corresponding observed trends were within the CESM ensemble range. There is a suggestion that the model produces higher interannual variability than what is observed. The trends between realizations of the same model differ strongly, which suggest that internal climate variability can strongly amplify or mask local trends in extreme events.