Pre-clinical models have an important role in cancer research. This review looks at the history as well as advantages and limitations of several animal models used in the study of pediatric brain tumors. The neoplasms of specific interest are the medulloblastoma/primitive neuroectodermal tumor category of brain tumors, because of their relevance to concurrent clinical research and practice. Given the historic poor record for direct heterotransplantation of these malignancies, both the conventional subcutaneous site and immune-privileged sites are assessed, with particular emphasis on models adapted to chemotherapeutic studies. The subcutaneous site is easiest to monitor but may be sub-optimal for therapeutic assays of brain tumors due to low engraftment rate, the absence of a blood-brain barrier equivalent and other lack of similarity to the clinical situation. The addition of a basement membrane attachment matrix (Matrigel) has been shown to enhance engraftment rate and cell yield at the subcutaneous site. In contrast, the intracerebral site, which is an area of immune-privilege, is biologically suitable for the study of brain tumors and their therapy, but there is no opportunity for sequential treatment courses or assessment of tumor growth prior to the death of the animal. An intraocular xenograft model has been shown to mimic human brain tumors in its resemblance of access to systemic agents to that via the blood-brain barrier and offers advantages over other methods for the pre-clinical study of chemotherapy in medulloblastoma/primitive neuroectodermal tumor. The intraocular site allows engraftment from very small clinical samples, is suitable for sequential treatment phases and provides easy access for monitoring of tumor progress. Chemotherapeutic agents which have been evaluated in various xenograft models, both as single agents and in combination, as well as other and novel approaches to the treatment of brain tumors are reviewed.