Objective: To compare the similarities and differences between untargeted and targeted metabolomics in the detection of biomarkers of gestational diabetes mellitus. Methods: Serum samples were collected from 20 patients with GDM and 20 normal pregnant women. Untargeted metabolomics and targeted metabolomics analysis were performed at the same time to compare the quality control results, metabolite detection and diagnostic efficacy of the two technologies. Results: The quality control of both techniques was good, and the detection system of untargeted metabolomics was more stable. The OPLS-DA model showed that the sample distribution trend of targeted metabolomics could better differentiate patients with GDM. Comparing the types of metabolites detected by the two methods, the former detected 639, mainly lipids, while the latter only detected 268, mainly amino acids and polypeptide analogues. Citric acid detected by targeted metabolomics had the highest diagnostic efficacy with an area under the curve of 0.983. Conclusion: Both technologies can screen out biomarkers with certain diagnostic efficiency, untargeted metabolomics cover a wider range of potential biomarkers, while targeted metabolomics can screen out biomarkers with higher diagnostic efficiency. The two have complementary advantages in the screening of biomarkers.