We assess the simulation of Australian mean climate and rainfall variability in a new version of the CSIRO coupled ocean-atmosphere global climate model (GCM). The new version, called Mark 3.6 (Mk3.6), differs from its recent predecessors (Mk3.0 and Mk3.5) by inclusion of an interactive aerosol scheme, which treats sulfate, dust, sea salt and carbonaceous aerosol. Other changes include an updated radiation scheme and a modified boundary-layer treatment. Comparison of the mean summer and winter climate simulations in Mk3.6 with those in Mk3.0 and Mk3.5 shows several improvements in the new version, especially regarding winter rainfall and sea-level pressure. The improved simulation of Australian mean seasonal climate is confirmed by calculation of a non-dimensional skill score (the 'M-statistic'), using data from all four seasons. However, the most dramatic improvement occurs in the model's simulation of the leading modes of annual rainfall variability, which we assess using empirical orthogonal teleconnections (EOTs). Compared to its predecessors and several international GCMs, Mk3.6 is best able to capture the spatial pattern of the leading rainfall mode, which represents variability due to the El Nino Southern Oscillation (ENSO). Mk3.6 is also best able to capture the spatial pattern of the second rainfall mode, which corresponds to increased rainfall in the northwest, and decreased rainfall over eastern Australia. We propose a possible mechanism for the improved simulation of rainfall variability in terms of the role of interactive dust in Mk3.6. By further suppressing convection over eastern Australia during El Nino events, dust feedbacks may enhance rainfall variability there, in tune with the model's ENSO cycle. This suggests that an interactive aerosol treatment may be important in a GCM used for the study of Australian climate change and variability. Mechanistic sensitivity studies are needed to further evaluate this hypothesis.