It has long been thought that reducing mean-state biases would lead to improvements in variability. However, so far, there is no confirmation of a relation between model mean biases and variability. While most coupled models exhibit substantial sea surface temperature (SST) biases in the Tropical Atlantic, they are still able to reproduce reasonable SST variability in the basin. We investigate the relationship between the first- and second-order moments of the SST distribution in the equatorial Atlantic using Coupled Model Intercomparison Project Phase 5 simulations. Results suggest that the ability of the coupled models to properly reproduce the interannual variability is linked to their ability to simulate a realistic seasonal cycle evolution, that is, a realistic cold tongue development and a realistic Bjerknes feedback during the beginning of summer, rather than to their ability to represent the summer SST climatology.