Computable general equilibrium (CGE) models of world trade, often presented as demonstrating the benefits of trade liberalization, now make much more modest forecasts than they did just a few years ago. The estimated benefits are not only small in the aggregate, but also skewed toward developed countries; the expected contribution of trade liberalization to economic development and poverty alleviation is extremely limited. Related calculations, for the expected benefits of services liberalization, trade facilitation measures, and long-term productivity gains from trade liberalization, remain problematical and/or speculative. The empirical limitations of CGE forecasts rest on broader theoretical weaknesses: the models are largely locked within a static framework, and remarkably assume that trade policy causes no changes in total employment, up or down. Models built on more adequate theories, which have only begun to appear, would paint a very different picture of the effects of trade liberalization.