This study focuses on the annual cycle of global ocean heat content and its variation with depth. Our primary objective is to evaluate a recent suite of coupled ocean-atmosphere simulations of the twentieth century in the context of available observations. In support of this objective, we extend the analysis and interpretation of observational estimates. In many respects, the collection of models examined compare well with observations. The largest signal in the annual cycle of ocean heat content is in the midlatitudes, where all the models do a credible job of capturing the amplitude, phasing, and depth penetration. Judging the models' performance at high latitudes is more complex because of the sparseness of observations and complications owing to the presence of sea ice. The most obvious problems identified in this study are in the tropics, where many climate models continue to have troublesome biases.