An attempt has been made to use paleoclimatic data from the last glacial maximum to evaluate the sensitivity of two versions of an atmosphere/mixed-layer ocean model. Each of these models has been used to study the C02-induced changes in climate. The models differ in their treatment of cloudiness, with one using a fixed cloud distribution and the other using a simple parameterization to predict clouds. The models also differ in the magnitude of their response to a doubling of atmospheric C02, with the variable cloud model being nearly twice as sensitive as the fixed cloud version. Given the distributions of continental ice sheets, surface albedo, and the reduced carbon dioxide concentration of the ice age, the climate of the last glacial maximum (LGM) is simulated by each model and compared with the corresponding simulation of the present climate. Both models generate differences in sea surface temperature and surface air temperature which compare favorably with estimates of the actual differences in temperature between the LGM and the present. However, it is difficult to determine which version of the model is more realistic in simulating the ice age climate for two reasons: 1) the differences between the two models are relatively small; and 2) there are substantial uncertainties in the pateoclimatic data. Neverthless, the similarity between the LGM simulations and the available paleoclimatic data suggests that the estimates of C02-induced climate change obtained from these models may not be too far from reality.