Glioblastoma is an aggressive brain tumour with a patient median survival of approximately 14 months. The development of innovative treatment strategies to increase the life span and quality of life of patients is hence essential. This requires the use of appropriate glioblastoma models for preclinical testing, which faithfully reflect human cancers. The aim of this study was to establish glioblastoma patient-derived xenografts (PDXs) by heterotopic transplantation of tumour pieces in the axillae of NMRI nude mice. Ten out of 22 patients' samples gave rise to tumours in mice. Their human origin was confirmed by microsatellite analyses, though minor changes were observed. The glioblastoma nature of the PDXs was corroborated by pathological evaluation. Latency times spanned from 48.5 to 370.5 days in the first generation. Growth curve analyses revealed an increase in the growth rate with increasing passages. The methylation status of the MGMT promoter in the primary material was maintained in the PDXs. However, a trend towards a more methylated pattern could be found. A correlation was observed between the take in mice and the proportion of Sox2+ cells (r = 0.49, p = 0.016) and nestin+ cells (r = 0.55, p = 0.007). Our results show that many PDXs maintain key features of the patients' samples they derive from. They could thus be used as preclinical models to test new therapies and biomarkers.