Gastric cancer is the world's third leading cause of cancer mortality. In spite of significant therapeutic improvements, the clinical outcome for patients with advanced gastric cancer is poor; thus, the identification and validation of novel targets is extremely important from a clinical point of view. We generated a wide, multilevel platform of gastric cancer models, comprising 100 patient-derived xenografts (PDX), primary cell lines, and organoids. Samples were classified according to their histology, microsatellite stability, Epstein–Barr virus status, and molecular profile. This PDX platform is the widest in an academic institution, and it includes all the gastric cancer histologic and molecular types identified by The Cancer Genome Atlas. PDX histopathologic features were consistent with those of patients' primary tumors and were maintained throughout passages in mice. Factors modulating grafting rate were histology, TNM stage, copy number gain of tyrosine kinases/KRAS genes, and microsatellite stability status. PDX and PDX-derived cells/organoids demonstrated potential usefulness to study targeted therapy response. Finally, PDX transcriptomic analysis identified a cancer cell–intrinsic microsatellite instability (MSI) signature, which was efficiently exported to gastric cancer, allowing the identification, among microsatellite stable (MSS) patients, of a subset of MSI-like tumors with common molecular aspects and significant better prognosis. In conclusion, we generated a wide gastric cancer PDX platform, whose exploitation will help identify and validate novel "druggable" targets and optimize therapeutic strategies. Moreover, transcriptomic analysis of gastric cancer PDXs allowed the identification of a cancer cell–intrinsic MSI signature, recognizing a subset of MSS patients with MSI transcriptional traits, endowed with better prognosis. © 2019 American Association for Cancer Research.