OBJECTIVE: To evaluate the predictability of virtual orthodontic setup technology by comparing it with posttreatment models in terms of ABO Model Grading System (ABO OGS) scores. MATERIALS AND METHODS: Posttreatment models and printed virtual setups made from the corresponding initial casts of 26 cases were used. The initial models were scanned and virtual setups were fabricated using a proprietary virtual setup software program. Evaluation of the printed models of the virtual setups and final mode ls was performed using the ABO OGS. RESULTS: We found statistically significant differences (p<0.001) between the posttreatment plaster models and the printed virtual setup models in total scores, marginal ridges (MR), buccolingual inclination (BLI), occlusal contacts (OC), occlusal relationship (OR), and interproximal contacts (IPC) (p<0.05). The only statistically significant correlations found were in alignment (A) and marginal ridges (MR) measurements although they were low positive correlations. CONCLUSIONS: The greatest difference between final treatment models and virtual setup models was in the OC due to overlapping of the occlusal surfaces in the virtual setup software. Printed models of virtual setups can be predictive in A and MR measurements. Overall, the 3-dimensional printed virtual setup models had better MR, BLI, and IPC, worse OC, OR, and total score, and equivalent A and OJ when compared to the posttreatment plaster models.