Background: Accurate diagnosis and risk stratification of adult-type diffuse gliomas remain hampered by limited neuroimaging specificity, invasive tissue biopsy, and low sensitivity of conventional liquid biopsies. Peripheral blood mononuclear cells (PBMCs) reflect tumor-induced systemic immune remodeling, representing a promising reservoir of epigenetic biomarkers.Methods: We profiled PBMC methylation from 559 individuals, including 152 gliomas, 15 brain metastases (BM), 33 primary central nervous system lymphomas (PCNSL), 194 extracranial tumors (ETs), and 165 healthy controls (HCs) using the Infinium MethylationEPIC BeadChip (935K). Machine learning models were built and validated via 8:2 stratified sampling with 50 iterations for glioma detection, differential diagnosis, subtyping, and grading. Univariate and multivariate Cox regression were applied to predict glioblastoma (GBM) prognosis.Findings: Glioma patients exhibited a predominant global hypomethylation pattern in PBMCs, with alterations enriched in T cell subsets. The model distinguishing gliomas from HCs/ETs achieved a median AUROC of 0.983. For differentiating gliomas from BM/PCNSL, the median AUROC was 0.959. Subtyping (GBM vs. oligodendroglioma [OD]/astrocytoma [AS]) reached a median AUROC of 0.923, and grade 4 vs. lower-grade yielded 0.855. Prognostic risk scores for progression-free survival (PFS) and overall survival (OS) in GBM remained independent prognostic factors after adjustment for TERT promoter mutation and MGMT promoter methylation (PFS: HR = 1.46, 95% CI: 1.30–1.63; OS: HR = 1.34, 95% CI: 1.13–1.60).Interpretation: This study defines a comprehensive PBMC methylation landscape of gliomas, identifying a subtype-specific systemic hypomethylation signature. Our epigenetic liquid biopsy enables noninvasive detection, classification, and prognostic stratification for gliomas.