Vibrio parahaemolyticus strains recovered from human diarrheal stools (one in 1975 and two in 2001) and environmental sources (four, between 2008 and 2010) were investigated for the presence of virulence genes (trh, tdh, and vpadF), pandemic markers (orf8, toxRSnew), and with respect to their pathogenic potential in two systemic infection models. Based only on the presence or absence of these genetic markers, they were classified as follows: the environmental strains were non-pathogenic, whereas among the clinical strains, the one isolated in 1975 was pathogenic (non-pandemic), and the other two were pathogenic (pandemic). The pathogenic potential of the strains was evaluated in mice and Galleria mellonella larvae infection models, and except for the clinical (pathogenic, non-pandemic) isolate, the others produced lethal infection in both organisms, regardless of their source, serotype, and genotype (tdh, orf8, toxRSnew, and vpadF). Based on mice and larval mortality rates, the strains were then grouped according to virulence (high, intermediate, and avirulent), and remarkably similar results were obtained by using these models: The clinical strain (pathogenic and non-pandemic) was classified as avirulent, and other strains (four non-pathogenic and two pandemic) were considered of high or intermediate virulence. In summary, these findings demonstrate that G. mellonella larvae can indeed be used as an alternative model to study the pathogenicity of V. parahaemolyticus. Moreover, they raise doubts about the use of traditional virulence markers to predict pathogenesis of the species and show that reliable models are indispensable to determine the pathogenic potential of environmental isolates considered non-pathogenic, based on the absence of the long-standing virulence indicators.