Abstract Single-cell studies have suggested heterogeneity of neoplastic cell states in diffuse gliomas. IDH-mutant astrocytoma and oligodendrogliomas (IDH-glioma) is described by cell states resembling including oligodendrocytes(OC), astrocytes (AC), and cycling neural progenitor cells (NPC); and IDH-wildtype glioblastoma (GBM-IDHwt) is marked by cell states resembling AC, NPC, oligodendrocyte progenitors (OPC) and mesenchymal (MES). Collective experience indicates that cell states are dynamic and can transition, at times recapturing brain development and inflammatory wound response. However, the spatial architecture of cell states, cell types and their interactions are yet to be investigated. We quantified the transcriptomics of 1,941 regions of interest (ROIs) from 19 IDH-glioma (non-1p/19q-codeleted n = 11, 1p/19q-codeleted n = 8) and 11 GBM-IDHwt tumors using the NanoString GeoMx Cancer Transcriptome Atlas (CTA, n = 1,811 genes) to study their spatial heterogeneity. Where appropriate, ROIs were selected to represent tumor core, infiltration edge, pseudopalisading necrosis, microvascular proliferation, and histological normal areas, if available. CTA profiling showed strong inter- and intra-tumor heterogeneity. Genes involved in oncogenesis and extra-cellular matrix such as VEGFA and VCAN, and macrophages (including CD74 and CCL3/1) showed the highest spatially variation. We identified substantial tumor immune environmental differences across ROIs including immune code, monocyte-like rich and lymphocyte-rich groups. The GBM-IDHwt cell states OPC and NPC were highly positively correlated (p-value = 5.65e-122) across ROIs suggesting spatial co-localization, but MES cell states were negatively correlated with the other three GBM cell states. M0 macrophage abundance was positively correlated with the MES state but negatively correlated with other cell states in GBM-IDHwt. The MES state was significantly lower in infiltrating edge of the tumor (p-value = 1.39e-34) but higher in the pseudopalisading necrosis region (p-value = 7.66e-30). Overall, this spatial transcriptomics dataset provides spatially informed relationships of neoplastic cell states and the immune microenvironment in diffuse glioma.