INTRODUCTION: Early diagnosis is crucial for improving the prognosis of patients with gastric cancer (GC). However, the currently used biomarkers for diagnosing GC have limited sensitivity and specificity. This study aimed to develop a novel diagnostic model based on miRNAs from glycosylated extracellular vesicles and evaluate its effectiveness in diagnosing gastric cancer. METHODS: GlyExo-capture technology was used to isolate glycosylated extracellular vesicles from serum samples. The signatures were screened in a discovery cohort of GC patients (n=55) and non-disease controls (n=46) using an integrated process, including high-throughput sequencing technology, screening using a complete bioinformatics algorithm, validation using RT-qPCR, and evaluation by constructing a diagnostic model. The diagnostic model was evaluated in an independent validation cohort (n=139). RESULTS: We developed a diagnostic model for GC based on five miRNA pairs. This diagnostic model demonstrated high sensitivity, specificity, and stable performance in distinguishing GC patients from non-cancer controls with AUC of 0.930 in the independent validation cohort, particularly in differentiating early-stage GC from benign patients. The markers also showed excellent performance in indicating perineural invasion status and lymph node metastasis in the testing cohort. DISCUSSION: The model demonstrated high sensitivity and specificity in diagnosing patients with GC, especially in differentiating early-stage GC from benign patients. The five miRNA pairs could also aid in making treatment decisions. Thus, miRNAs derived from glycosylated exosomes are promising biomarkers for cancer diagnosis.