Swarm intelligence refers to a collective intelligent behavior that emerged from the interactions of individuals obeying simple behavioral rules, and labor division is one of its most profound features. According to the individual-individual interactions and individual-environment interactions, we present a framework description of labor division in swarm intelligence and analyzes its characteristics of individual specialization, role plasticity, and self-organization. From the perspective of self-organization, four kinds of labor division models, namely, activator-inhibitor, stimulus-response, individual sorting, and foraging for work, are compared and analyzed in detail, with the aim of concluding the general modeling regularities for self-organized labor division. As well as the applications of self-organized labor division, its application range, and solving methods are further analyzed. Finally, some perspectives on the development of self-organized labor division in swarm intelligence are drawn from the mechanisms and models of labor division, as well as the solving methods for allocation problems and optimization problems.