A common assumption in studies of multispecies mutualism is that a single partner species provides the best services, while other species range from lower-quality mutualists to outright cheaters. In contrast, we find that Acacia drepanolobium trees in north central Kenya benefit from recurring transitions in occupancy among four ant mutualists.  To investigate lifetime fitness effects of variation in ant partners, we used a long-term, large-scale data set to parameterize demographic models for acacias in which growth, reproduction, and survival are size- and ant species-dependent, and the transitions between the four ant species on a tree are dependent on tree size and the current ant occupants. This model provides highly robust predictions of the observed size x ant species distributions of acacias. By simulating the removal of one or more ants from the system, we investigated the effects of different ant communities on tree fitness. Having two or more ant mutualists generally increases fitness over having only a single species of ant occupant, even when that single species is the most aggressive and protective mutualist. Most strikingly, the ant species that dramatically reduces tree survival, but that also increases fruiting rates, actually benefits acacias when existing in combination with other species, due to the strong competitive hierarchy between the ants, and the effects of tree size on these interactions. Overall, our results suggest that sets of competing mutualists may strongly benefit longer-lived partners that are able to reap contrasting benefits from different mutualist partners over their lifetimes.