The “wisdom of the crowds” refers to the idea that the aggregated performance of a group of people on a challenging task may be superior to the performance of any of the individuals. For some tasks, like estimating a single quantity, it is straightforward to aggregate individual behavior. For more complicated multidimensional or sequential tasks, however, it is not so straightforward. Cognitive models of behavior are needed, to infer what people know from how they behave, and allow aggregation to be done on the inferred knowledge. We provide a case study of this role for cognitive modeling in the wisdom of crowds, using a multidimensional sequential optimization problem, known as the bandit problem, for which there are large differences in individual ability. We show that, using some established cognitive models of people’s decision-making on these problems, aggregate performance approaches optimality, and exceeds the performance of the vast majority of individuals.