We explore alternative approaches to numerical solutions oflarge rational-expectations models. We discuss and compare severalcurrent alternatives, focusing on the trade-offs in accuracy, space,and speed. The models range from representative-agent models withmany goods and capital stocks, to models of heterogeneous agents withcomplete or incomplete asset markets. The methods includeperturbation and projection methods. We show that these methods arecapable of analyzing moderately large models even when we use onlyelementary, general-purpose numerical methods.