Today we experience an accelerated melting of sea ice in the Arctic which global circulation models (GCMs) do not adequately predict. We believe that one of the reasons for this may be shortcomings in the description of the energy balance in current GCMs, as previous studies have shown that today’s GCMs are unable to capture the annual cycle of sea-ice albedo, particularly in summer when the models overestimate the albedo. Here we investigate a physically based sea-ice albedo description for the ECHAM5 GCM, which separates between snow-covered sea ice, bare sea ice, melt ponds and open water (separately for the albedos and the areal fractions). The new albedo scheme includes important components such as albedo decay due to snow ageing, bare sea-ice albedo dependent on ice thickness and a melt-pond albedo dependent on melt-pond depth. The explicit treatment of melt-pond albedos has, to our knowledge, not been included in a GCM before and represents a substantial improvement when simulating the annual cycle of the sea-ice albedo. The new algorithm performs well in modeling the distribution of melt ponds: the timing of their formation and their extent agree well with observations. Simulations showed albedo reductions for shallow ponds on first-year ice to be larger than reductions for deeper ponds on multi-year ice, because the pond areal fraction is larger. This indicates that melt ponds may play an even more important role in ice melt in the Arctic in the years to come, as more and more of the rough multi-year sea ice is being replaced by the smooth first-year sea ice. The reduced sea-ice albedo leads to overall reduced sea-ice thickness, concentration and volume, but with temporal and spatial variations. But in August the pattern is spatially homogeneous, with reduced albedo, thickness and concentrations for all areas where the new scheme has a significant effect.