Amplification of the MET oncogene occurs in 2–4% of gastroesophageal cancers and defines a small and aggressive subset of tumors. Although in vitro studies have given very promising results, clinical trials with MET inhibitors have been disappointing, showing few and short lasting responses. The aim of the work was to exploit a MET-amplified patient-derived xenograft model to optimize anti-MET therapeutic strategies in gastroesophageal cancer. We found that despite the high MET amplification level (26 gene copies), in the absence of qualitative or quantitative alterations of EGFR, MET inhibitors induced only tumor growth inhibition, whereas dual MET/EGFR inhibition led to complete tumor regression. Importantly, the combo treatment completely prevented the onset of resistance, which quite rapidly appeared in tumors treated with MET monotherapy. We found that this secondary resistance was due to EGFR activation and could be overcome by dual MET/EGFR inhibition. Similar results were also obtained in a MET-addicted, established gastric cancer cell line. In vitro experiments performed on tumor-derived primary cells confirmed that MET inhibitors were not able to abrogate the activation of downstream transducers and that only the combined MET/EGFR treatment completely shut off the signaling. Previously reported cases, as well as those described here, showed only partial and transient sensitivity to anti-MET therapy. The finding that combined anti-MET/EGFR therapy—even in the absence of EGFR genetic alterations—induced complete and durable response, represents a proof of concept and guarantees further investigations, opening a new perspective of treatment for these patients.