To evaluate the clinical predictive value of serum alpha-fetoprotein variants (AFP-L2, AFP-L3) in combination with maternal serum prenatal screening biomarkers in predicting fetal trisomy 21 and trisomy 18. We analyze the data of singleton pregnant women at 15–20+6 weeks of 731,922 gravidas from October 2007 to September 2019. The research objects were separated into the following groups: control (n = 569), trisomy 21 (n = 116), and trisomy 18 (n = 52). The cases were diagnosed by chromosomal karyotypic analysis of amniotic fluid cells. Level of AFP-L2 and AFP-L3 were detected in maternal serum among control women and patients. Receiver operator characteristic analysis, detection rate, false positive rate, false negative rate, positive predictive value, negative predictive value, positive likelihood ratio and negative likelihood ratio, comprehensive discriminant improvement, net weight classification improvement, decision curve analysis and Hosmer–lemeshow (H-L) test were used to investigate the predictive value of free β-hCG, AFP, AFP-L2 and AFP-L3 on the risk models of trisomy 21, 18. There was a statistically significant difference in maternal serum AFP-L2 and AFP-L3 multiple of the median (MoM) among the trisomy 21, trisomy 18, and control groups. The AUCs of AFP-L2 and AFP-L3 for the screening trisomy 21 and trisomy 18 fetus were 0.785, 0.758 and 0.775, 0.754. According to ROC, the optimal cut-off values of AFP-L2 and AFP-L3 for predicting trisomy 21 and trisomy 18 fetuses all were 1.09 MoM and 1.30 MoM, respectively. The risk-calculation model constructed by AFP-L2 + AFP-L3 MoM manifested better efficiency than the original single-value truncation method using AFP MoM alone. Compared with different modeling methods, the AUC of trisomy 21 fetuses predicted by AFP-L2 + AFP-L3 + free β-hCG achieved an optimal value (0.938), while the AUC of trisomy 18 fetus predicted by AFP-L2 + free β-hCG was the best (0.991). Compared with AFP, the IDI of AFP-L2 or AFP-L3 alone increased 9.56% and 12.34%; the NRI increased 26.50% and 26.70 in predicting trisomy 21. For trisomy 18, the IDI of AFP-L2 or AFP-L3 alone declined with 8.12% and 1.52%; the NRI declined with 13.84% and 8.54%. In the combined model, the model with best detection rate, false positive rate and positive likelihood ratio was AFP-L2 + AFP-L3 + free β-hCG, followed by AFP-L2 + free β-hCG and AFP-L3 + free β-hCG, and finally AFP + free β-hCG. Maternal serum AFP-L2 and AFP-L3 in the second trimester is a good marker for screening trisomy 21 and trisomy18 with high sensitivity and specificity. The combined screening results are better than the single marker, and the efficiency of AFP-L2 + AFP-L3 + free β-hCG is the best.