Considering its relationships with real life and mathematics, mathematical modeling is one of the most effective ways of learning and teaching mathematics. In this study, the effect of teaching with mathematical modeling on students' mathematics achievement was examined using the meta-analysis method. An effect size of 51 was achieved with 45 studies that met the inclusion criteria. Random effects model and Hedges's g value were used to calculate effect sizes, and the source of heterogeneity was tried to be determined through moderator analyses. Possible publication bias was examined with the funnel plot, Rosenthal's fail-safe N, Orwin's fail-safe N, Egger's linear regression analysis, Begg and Mazumdar's rank correlation, and Duval and Tweedie's trim-and-fill methods, and no findings of publication bias were found. According to the effect size results showed that the effect of teaching with mathematical modeling on students' mathematics achievement is high level (g=0.845, p<0.001, 95% CI:0.712-0.978). In addition, the effect sizes show a heterogeneous distribution (Q=174.533, df=50, p<0.001). While the effect of teaching with mathematical modeling on students' mathematics achievement differed statistically according to the moderators of research design (Q[subscript B]=11.894, df=1, p<0.05) and the contents of mathematical modeling (i.e., learning area/subject) (Q[subscript B]=19.941, df=6, p<0.05), it did not differ statistically according to the moderators of publication type (Q[subscript B]=3.666, df=2, p>0.05), level of education (Q[subscript B]=3.497, df=3, p>0.05), or implementation period (Q[subscript B]=3.200, df=2, p>0.05). Some suggestions are presented at the end of the study for researchers and practitioners regarding teaching with mathematical modeling.