Accumulating evidence demonstrates that defining molecular subtypes based on objective genetic alterations may permit a more rational, patient-specific approach to molecular targeted therapy across various cancers. The objective of this study was to subtype primary glioblastoma (pGBM) based on MicroRNA (miRNA) profiling in Chinese population. Here, miRNA expression profiles from 82 pGBM samples were analyzed and 78 independent pGBM samples were used for qRT-PCR validation. We found that two distinct subgroups with different prognosis and chemosensitivities to temozolomide (TMZ) in Chinese pGBM samples. One subtype is TMZ chemoresistant (termed the TCR subtype) and confers a poor prognosis. The other subtype is TMZ-chemosensitive (termed the TCS subtype) and confers a relatively better prognosis compared with the TCR subtype. A classifier consisting of seven miRNAs was then identified (miR-1280, miR-1238, miR-938 and miR-423-5p (overexpressed in the TCR subtype); and let-7i, miR-151-3p and miR-93 (downregulated in the TCR subtype)), which could be used to assign pGBM samples to the corresponding subtype. The classifier was validated using both internal and external samples. Meanwhile, the genetic alterations of the TCR and TCS subtypes were also analyzed. The TCR subtype was characterized by no IDH1 mutation, and EGFR and Ki-67 overexpression. The TCS subtype displayed the opposite situation. Taken together, the results indicate a distinct subgroup with poor prognosis and TMZ-chemoresistance.