Little is known about the mental representation of large multidigit numbers that are usually beyond our personal experience. The present study explored the processing mechanisms of these numbers in a series of experiments, using the numerical comparison task. Experiment 1 included within and between-scale comparisons of multidigit numbers varying in their left digits (e.g., 8,000,000), with one group comparing small numbers (tens, hundreds, and thousands) and the other large ones (millions, billions, and trillions). In Experiment 2, comparisons of small (tens, hundreds) and large (millions, billions) multidigit numbers that varied in their left and right digits (e.g., 8,000,003) were presented in separate blocks. Experiment 3 presented small and large multidigit numbers (from tens to trillions) that varied in their left digits in the same block. We found novel compatibility effects between the left digit and scale components, as well as between the left digit, right digit, and scale components, and extended the previously reported unit-decade compatibility effect to larger scales. We also obtained global and scale distance effects for all scales in most conditions. Both compatibility and distance effects showed context dependency in large, but not small, multidigit numbers. Overall, these results demonstrate that small and large multidigit numbers are processed differently. We discuss these differences and propose a processing model that accounts for them.