Despite the importance and wide exploitation of heterosis in commercial crop breeding, the molecular mechanisms behind this phenomenon are not completely understood. Recent studies have implicated changes in DNA methylation and small RNAs in hybrid performance, however, it remains unclear whether epigenetic changes are a cause or consequence of heterosis. Here, we analyze a large panel of over 500 A. thaliana epigenetic hybrid plants (epiHybrids), which we derived from near-isogenic but epigenetically divergent parents. This proof-of-principle experimental system allowed us to quantify the contribution of parental methylation differences to heterosis. We measured traits such as leaf area (LA), growth rate (GR), flowering time (FT), main stem branching (MSB), rosette branching (RB) and final plant height (HT) and observed several strong positive and negative heterotic phenotypes among the epiHybrids. Using an epigenetic quantitative trait locus mapping approach, we were able to identify specific differentially methylated regions (DMRs) in the parental genomes that are associated with hybrid performance. Sequencing of methylomes, transcriptomes and genomes of selected parent-epiHybrid combinations further showed that these parental DMRs most likely mediate remodeling of methylation and transcriptional states at specific loci in the hybrids. Taken together, our data suggest that locus-specific epigenetic divergence between the parental lines can directly or indirectly trigger heterosis in Arabidopsis hybrids independent of genetic changes. These results add to a growing body of evidence that points to epigenetic factors as one of the key determinants of hybrid performance.