The sea-ice performance of two versions of the Australian Community Climate and Earth System Simulator (ACCESS-CM), ACCESS1.0 and ACCESS1.3, participating in the Coupled Model Inter-comparison Project phase 5 (CMIP5) experiment, is assessed. Before completing the CMIP5 experiment, detailed testing was undertaken to improve the representation of sea ice. We determined optimal parameter values related to the sea-ice physics to produce as realistic a global sea-ice distribution as possible². These optimal values were selected based on the closest match between the model and observationally constrained model climatologies. The sea-ice distribution shows a similar degree of sensitivity to parameters determining ice-ocean stress, mechanical redistribution, oceanic heat and shortwave radiation. The sensitivity of ice volume is stronger than that of ice area indicating that the parameters mostly influence the ice thickness. The performance of the ocean model is crucial in producing a realistic sea-ice cover. After completing the CMIP5 experiment, comparisons with model output from other international climate modelling centres were also included in the assessment process¹.  The assessment took into account modelled climatologies and interannual variability of the sea-ice extent, concentration, thickness and transport.  We found that the ACCESS-CM models give good simulations of the global sea ice, within the scatter of models studied, and is one of the top performing models for sea-ice metrics. In the Arctic, the ACCESS-CM models generally capture the observed decline of sea ice over the period of 1981-2011, although the sea-ice reductions occur in the Laptev and Kara Seas rather than in the Chukchi and Beaufort Seas as observed. In the Antarctic, the simulated changes in the sea ice over the period of 1981-2011 are generally smaller than in the Arctic, but the ACCESS-CM models do not capture the observed small increase in sea ice. This observed increase, however, is likely to occur due to the internal variability of the climate system which is supported by the fact that the ACCESS-CM CMIP5 simulations contain several 30-year periods where the Antarctic sea ice increases with patterns resembling the observed one.