Given the global biodiversity crisis, it is crucial to identify methods best suited for conducting inventories. We evaluated the relative merits of Malaise traps (MT) and ground-level yellow pan traps (YPT) to sample male velvet ants (Hymenoptera: Mutillidae) in a Neotropical savanna biodiversity hotspot. We compared richness, number of captures, evenness, composition, and body size of male velvet ants (Hymenoptera: Mutillidae) sampled with both methods in four sites at Parque Nacional da Chapada dos Veadeiros, central Brazil, during 19 d. We expected reduced diversity and smaller body size of velvet ants sampled with YPT, because they target visually oriented insects that are active closer to the ground, whereas MT represent a passive method that intercepts insects flying at different heights. Richness, total number of captures, and evenness of species and genera were significantly higher in MT. The body size of velvet ants captured with MT was significantly larger than those found in YPT. Generalized linear model and nonmetric multidimensional scaling analyses revealed a clear difference in the patterns of abundances and composition of velvet ants sampled with MT and YPT, especially for the genera Darditilla, Traumatomutilla, Lomachaeta, Pseudomethoca, Tallium, and Ephuta. YPT were effective at capturing few species that were rare in MT but, overall, MT were much more effective than YPT. We found similar patterns when using either species or genus for assessing samples obtained with MT or YPT, suggesting that ecological studies on Neotropical velvet ants may not require taxonomic resolution to the species level.