Three newly discovered Melanogaster species, namely M. cyaneus, M. diqingensis, and M. truncatisporus, are introduced and illustrated based on both morphological and molecular data from Sichuan and Yunnan provinces in China. A multigene phylogenetic analysis (nrITS, nrLSU, and rpb2) was performed mainly to verify the placement of the new species in Melanogaster. A second, nrITS-only phylogenetic analysis comprising more Melanogaster species for which only ITS sequences were available, was used to infer the relationship between the new species and as many known Melanogaster species as possible. Specimens of M. cyaneus, M. diqingensis, and M. truncatisporus formed three independent clades in a phylogenetic tree inferred from the ITS data set. The robust support from ITS for these clades and genetic similarity with other species being lower than 93.2% suggest that these three species are indeed distinct from the other Melanogaster species in the phylogeny. Morphologically, M. cyaneus is characterized by its blue or bluish gleba, light brown to yellowish brown peridium, and subglobose to globose basidiospores, 6.2–15 × 4.6–9.0 μm. Melanogaster diqingensis is distinguished from other Melanogaster species by its pale yellow to brown-yellow peridium and obovate to subglobose basidiospores, 3.0–5.1 × 2.0–4.0 μm. Melanogaster truncatisporus is diagnosed by its subglobose to globose or irregularly elongate-pyriform basidiomata, pale yellow to deeply orange-yellow peridium, and subglobose to globose or pyriform, truncate basidiospores. Additionally, infrageneric classification based on the number of peridium layers, the average thickness of the peridium, and the average length and width of basidiospores was tested with M. cyaneus, M. diqingensis, and M. truncatisporus. Orthogonal partial least squares discriminant (OPLS-DA) analysis placed the three new species within the Melanogaster, Rivulares, and Variegati sections, respectively. However, the morphologically circumscribed sections were not monophyletic in the phylogenetic tree. Therefore, the current infrageneric classification should be abandoned.