The successive reports of the Intergovernmental Panel on Climate Change (IPCC) illustrate an apparent contradiction. On the one hand, the large scale climatic change in response to the increase of greenhouse gases is structured following patterns which have remained surprisingly stable throughout the development of climate models. Already in the 1980s model simulations of climate change were characterized by a larger warming in polar areas and over the continents, and a tendency for precipitations to accentuate existing contrasts, with a drier climate in semiarid regions and more precipitations at mid-latitudes or near the Equator. On the other hand, models have made little progress in predicting more unanimously and more reliably the global amplitude of climate changes and their geographical distributions. This lack of progress is certainly linked with the role of the atmospheric dynamics in shaping up certain aspects of climate response, either small scales which affect atmospheric stratification, or synoptic scales, whose inherent complexity and nonlinear interactions also limit the possibility of more accurate regional predictions.