The development of novel therapies against melanoma would benefit from individualized tumor models to ensure the rapid and accurate identification of biomarkers of therapy response. Previous studies have suggested that patient-derived xenografts (PDXes) could be useful. However, the utility of PDXes in guiding real-time treatment decisions has only been reported in anecdotal forms. Here tumor biopsies from patients with stage III and IV metastatic malignant melanoma were transplanted into immunocompromised mice to generate PDXes. 23/26 melanoma biopsies generated serially transplantable PDX models, and their histology, mutation status and expression profile resembled their corresponding patient biopsy. The potential treatment for one patient was revealed by an in vitro drug screen and treating PDXes with the MEK inhibitor trametinib. In another patient, the BRAF mutation predicted the response of both the patient and its corresponding PDXes to MAPK-targeted therapy. Importantly, in this unselected group of patients, the time from biopsy for generation of PDXes until death was significantly longer than the time required to reach the treatment phase of the PDXes. Thus, it could be clinically meaningful to use this type of platform for melanoma patients as a pre-selection tool in clinical trials.