Gastric cancer displays marked molecular heterogeneity with aggressive behavior and treatment resistance. Therefore, good in vitro models that encompass unique subtypes are urgently needed for precision medicine development. Here, we have established a primary gastric cancer organoid (GCO) biobank that comprises normal, dysplastic, cancer, and lymph node metastases (n = 63) from 34 patients, including detailed whole-exome and transcriptome analysis. The cohort encompasses most known molecular subtypes (including EBV, MSI, intestinal/CIN, and diffuse/GS, with CLDN18-ARHGAP6 or CTNND1-ARHGAP26 fusions or RHOA mutations), capturing regional heterogeneity and subclonal architecture, while their morphology, transcriptome, and genomic profiles remain closely similar to in vivo tumors, even after long-term culture. Large-scale drug screening revealed sensitivity to unexpected drugs that were recently approved or in clinical trials, including Napabucasin, Abemaciclib, and the ATR inhibitor VE-822. Overall, this new GCO biobank, with linked genomic data, provides a useful resource for studying both cancer cell biology and precision cancer therapy. Display omitted •Living biobank includes 17 normal and 46 gastric cancer organoid lines•Organoid biobank encompasses most of the known molecular subtypes of gastric cancer•Organoids recapitulate the genomic and transcriptomic features of original tumors•High-throughput screen revealed potential target drugs for personalized therapy Leung and colleagues established a biobank of patient-derived gastric cancer organoids that encompasses a diverse array of subtypes and maintained long-term similarity to the original tumors. They used the organoids to perform large-scale drug screening that identified potential target drugs and could guide patient drug selection.