Abstract Intratumoral heterogeneity in glioblastoma is known and is considered a key feature of GBM’s clinical aggressiveness. To more fully characterize this phenomenon, we conducted methylation profiling on the EPIC array for 79 regions from 20 glioblastomas. Using the Heidelberg classifier, each of the 20 tumors had at least 1 region with a match to a GBM subtype with >0.84 score. Among these, 6 tumors (30%) showed homogeneous-high score matches to the same GBM subclass, while the remaining 14 tumors (70%) showed heterogeneity in GBM subclass, with either lower scores, or disparate matches to different glioma subtypes. Comparing these subclass-homogeneous versus subclass-heterogeneous groups of tumors using methylCIBERSORT deconvolution, the subclass-heterogeneous cases showed a significantly higher proportions of CD8+ T cells and microglia, and significantly lower proportions of CD4+ T cells, B cells and monocytes. While low tumor purity can be a cause of lower confidence scores in the methylation classifier, we did not find that the subclass-heterogeneous tumors had lower tumor purity compared to subclass-homogeneous cases, suggesting that this heterogeneity was due to factors other than tumor purity. Within the subclass-heterogeneous tumors we identified the common intratumorally most variable probes and examined their characteristics. Compared to the overall distribution of probes on the array, these intratumoral-variable probes were significantly more likely to be intergenic and located on enhancers. Conversely, they were significantly less likely to be on gene promoters and CpG islands. Gene ontology analysis showed that genes associated with these probes were enriched for pathways involved in DNA repair. Overall, our multisampling methylation profiling analysis identifies specific biologic correlates of intratumoral heterogeneity that is not simply due to tumor purity. Methylation changes associated with intratumoral heterogeneity converge on specific genomic patterns that warrant further study.