In our previous study, we demonstrated that four microRNAs (miRNAs) (miR-26a, miR-142-3p, miR-148a, and miR-195) that were downregulated in both plasma and tumor tissues were confirmed to be promising non-invasive diagnostic biomarkers for gastric cancer (GC). We used the quantitative reverse transcription polymerase chain reaction to assess the expression levels of the four miRNAs from paraffin-embedded surgical specimens of GC patients. Kaplan-Meier curves and log-rank test were applied to predict the correlation between miRNAs and cumulative overall survival (OS) of patients with GC. Besides, we performed in vitro assays including cell proliferation, migration, invasion and colony formation, and apoptosis. The median of miRNA expression in paraffin-embedded tissues were used as the cutoff value to classify patients into high or low expression groups. Down-regulation of miR-26a and miR-148a was significantly associated with shorter OS of GC patients either in the test set (miR-26a: P = 0.009; miR-148a: P = 0.005) or the validation set (miR-26a: P = 0.011; miR-148a: P = 0.024). When two sets were combined, Cox regression analysis demonstrated that both of miR-26a and miR-148a were independent prognostic factors for predicting OS of patients with GC (miR-26a: HR = 0.76, 95% CI = 0.61-0.94; miR-148a: HR = 0.73, 95% CI = 0.58-0.91). Furthermore, elevated expression of miR-26 significantly suppressed cell proliferation, migration, invasion and colony formation, and induced apoptosis of MGC-803 cells compared with negative control groups (P < 0.05). These findings supported miR-26a and miR-148a could serve as potential prognostic biomarkers for GC.