Although hyperparasitism frequently occur in parasitic insects, many aspects of this strategy remain unknown. We investigated possible fitness costs of hyperparasitism as influenced by host size. Our study was conducted with the facultative hyperparasitoid Pachycrepoideus dubius Ashmead (Hymenoptera: Pteromalidae), which parasitizes host species differing greatly in size. We compared some fitness traits (level of successful parasitism, development time, sex ratio and offspring size) of P. dubius developing on large secondary/primary (Delia radicum L. (Diptera: Anthomyiidae)/Trybliographa rapae Westwood (Hymenoptera: Figitidae)) or small secondary/primary host species (Drosophila melanogaster L./Asobara tabida Nees (Hymenoptera: Braconidae)). In no-choice and choice experiments, P. dubius was able to develop on different stages of T. rapae (L2 (endophagous), L4 (ectophagous), and pupae) but that it preferred to parasitize unparasitized D. radicum pupae over pupae parasitized by T. rapae. Furthermore, in P. dubius, hyperparasitism was associated with fitness costs (lower level of successful parasitism, smaller adult size) and these costs were greater on the smallest host complex. We hypothesize that the size of D. melanogaster pupae parasitized by A. tabida may be close to the suboptimal host size for P. dubius beneath which the costs of hyperparasitism make this strategy nonadaptive. Hyperparasitism in terms of trade-offs between host quality and abundance of competitors is discussed.