Standard therapy of osteosarcoma (OS) and Ewing sarcoma (EW) rests on cytotoxic regimes, which are largely unsuccessful in advanced patients. Preclinical models are needed to break this impasse. A panel of patient-derived xenografts (PDX) was established by implantation of fresh, surgically resected osteosarcoma (OS) and Ewing sarcoma (EW) in NSG mice. Engraftment was obtained in 22 of 61 OS (36%) and 7 of 29 EW (24%). The success rate in establishing primary cell cultures from OS was lower than the percentage of PDX engraftment in mice, whereas the reverse was observed for EW; the implementation of both in vivo and in vitro seeding increased the proportion of patients yielding at least one workable model. The establishment of in vitro cultures from PDX was highly efficient in both tumor types, reaching 100% for EW. Morphological and immunohistochemical (SATB2, P-glycoprotein 1, CD99, caveolin 1) studies and gene expression profiling showed a remarkable similarity between patient's tumor and PDX, which was maintained over several passages in mice, whereas cell cultures displayed a lower correlation with human samples. Genes differentially expressed between OS original tumor and PDX mostly belonged to leuykocyte-specific pathways, as human infiltrate is gradually replaced by murine leukocytes during growth in mice. In EW, which contained scant infiltrates, no gene was differentially expressed between the original tumor and the PDX. A novel therapeutic combination of anti-CD99 diabody C7 and irinotecan was tested against two EW PDX; both drugs inhibited PDX growth, the addition of anti-CD99 was beneficial when chemotherapy alone was less effective. The panel of OS and EW PDX faithfully mirrored morphologic and genetic features of bone sarcomas, representing reliable models to test therapeutic approaches.