Batesian mimics—benign species that receive protection from predation by resembling a dangerous species—often occur with multiple model species. Here we examine whether geographic variation in the number of local models generates geographic variation in mimic-model resemblance. In areas with multiple models, selection might be relaxed or even favour imprecise mimicry relative to areas with only one model. We test the prediction that model-mimic match should vary with the number of other model species in a broadly-distributed snake mimicry complex where a mimic and a model co-occur both with and without other model species. We found that the mimic resembled its model more closely when they were exclusively sympatric than when they were sympatric with other model species. Moreover, in regions with multiple models, mimic-model resemblance was positively correlated with the resemblance between the model and other model species. However, contrary to predictions, free-ranging natural predators did not attack artificial replicas of imprecise mimics more often when only a single model was present. Taken together, our results suggest that multiple models might generate a geographic mosaic in the degree of phenotype matching between Batesian mimics and their models.