Economic deliberations often require sophisticated analysis, but the behavioral algorithms and neurobiological mechanisms for complex deliberations remain unknown. We asked nonhuman primates (NHPs) to select optimal, or suboptimal, item subsets in a modified optimization problem. Surprisingly, the NHP behavior reflected computational algorithms – some that rely on combinatorial reasoning – and the deliberation times reflected the number of operations the best-matching algorithms used. Artificial neural networks revealed that constructing combinatorial solutions required greater computational resources than their greedy counterparts. These results provide mechanistic insights into the sophisticated reasoning skills that support economic deliberation.