The purpose of this study was to examine advanced diffusion-weighted magnetic resonance imaging (DW-MRI) models for differentiation of low- and high-grade tumors in the diagnosis of pediatric brain neoplasms. Sixty-two pediatric patients with various types and grades of brain tumors were evaluated in a retrospective study. Tumor type and grade were classified using the World Health Organization classification (WHO I-IV) and confirmed by pathological analysis. Patients underwent DW-MRI before treatment. Diffusion-weighted images with 16 b-values (0-3500 s/mm2) were acquired. Averaged signal intensity decay within solid tumor regions was fitted using two-compartment and anomalous diffusion models. Intracellular and extracellular diffusion coefficients (Dslow and Dfast), fractional volumes (Vslow and Vfast), generalized diffusion coefficient (D), spatial constant (μ), heterogeneity index (β), and a diffusion index (index_diff = μ × Vslow/β) were calculated. Multivariate logistic regression models with stepwise model selection algorithm and receiver operating characteristic (ROC) analyses were performed to evaluate the ability of each diffusion parameter to distinguish tumor grade. Among all parameter combinations, D and index_diff jointly provided the best predictor for tumor grades, where lower D (p = 0.03) and higher index_diff (p = 0.009) were significantly associated with higher tumor grades. In ROC analyses of differentiating low-grade (I-II) and high-grade (III-IV) tumors, index_diff provided the highest specificity of 0.97 and D provided the highest sensitivity of 0.96. Multi-parametric diffusion measurements using two-compartment and anomalous diffusion models were found to be significant discriminants of tumor grading in pediatric brain neoplasms.