Glioblastoma multiforme (GBM) is the most malignant brain tumor with a poor prognosis. A molecular level classification of GBM can provide insight into accurate patient-specific treatment. Competitive endogenous RNAs (ceRNAs), such as long non-coding RNAs (lncRNAs), play an essential role in the development of tumors and are associated with survival. However, the pattern of lncRNA-mediated ceRNA (LMce) crosstalk in different GBM subtypes is still unclear. In this study, we present a computational cascade to construct LMce networks of different GBM subtypes and investigate the lncRNA-mRNA regulations among them. Our results showed that although most lncRNAs and mRNAs in the different GBM subtype networks were the same, the regulation relationships of these RNAs were different among subtypes. 42.5%, 50.9%, 43.5% and 65.0% lncRNA-mRNA regulatory pairs were classic (CL)-, mesenchymal (MES)-, proneural (PN)- and neural (NE)-specific. In addition, our study identified 61, 132, 24 and 16 modules in which lncRNAs and mRNAs synergically competed with each other for miRNAs as CL-, MES-, PN- and NE-specific. CL- and MES-specific modules were mainly involved in biological functions such as cell proliferation, apoptosis and migration, while PN- and NE-specific modules were mainly related to DNA damage and cell cycle dysregulation. Survival analysis demonstrated that some modules could be potential prognostic markers of patients of CL and MES subtypes. This study uncovered the LMce interaction patterns in different GBM subtypes, identified subtype-specific modules with distinct biological functions, and revealed the potential prognostic markers of patients of different GBM subtypes. These results might contribute to the discovery of the GBM prognostic biomarkers and development of a more accurate therapeutic process.