South American savannas are a disjunct biome with an unclear evolutionary history. We tested hypotheses about their Quaternary history and evolution of savanna cores through fragmentation or dispersal from the Cerrado. We used genomic data (genotyping-by-sequencing) and ecological niche models of the Burnished-buff Tanager (Stilpnia cayana Linnaeus 1766) to evaluate intraspecific differentiation, gene flow, past range shifts, and landscape-genomics association. We found clear genomic differences between populations on each side of the Amazon basin and high admixture in the Marajó Island and Bolivia. Landscape genomics analysis indicated that the Amazon River, isolation by distance and temperature predict genomic differentiation in this bird. Taken together, the results suggest that a combination of dispersal from the Cerrado, isolation due to geographic distance and the Amazon River basin, and local adaptation shaped the species diversification. We propose considering three subspecies of S. cayana as full species, with one possibly of hybrid origin (S. huberi).