The aim of this study is to evaluate the performance of VEGF and endostatin levels in the differential diagnosis of malignant and benign ascites. The study included 101 consecutive patients with malignant ascites (55.2 ± 15.8 years, 63 men and 38 women) and 81 patients with benign ascites (53.0 ± 17.2 years, 51 men and 30 women). VEGF and endostatin levels in serum and ascites were determined by a sandwich enzyme immunoassay technique using a commercially available assay kit. The serum VEGF, ascites VEGF, and ascites endostatin levels of patients with malignant ascites were significantly higher than those in patients with benign ascites (P < 0.001), but there was no difference in serum endostatin levels between the two groups (P = 0.267). Ascites endostatin levels correlated positively with ascites VEGF (r = 0.5193, P < 0.01), and serum endostatin showed a low correlation with serum VEGF (r = 0.3291, P < 0.01) in patients with malignant ascites. Areas under the ROC curves of ascites VEGF, ascites endostatin, serum VEGF, and serum endostatin were 0.890, 0.815, 0.694, and 0.552, respectively. The combination of ascites VEGF and endostatin improved the sensitivity up to 90.1%, the specificity up to 87.7%, and the accuracy up to 89.0% in the differential diagnosis of malignant and benign ascites. VEGF and endostatin levels in ascites appear to be suitable for differentiating between malignant and benign ascites, which can be applied to clinical examination.