Purpose: The aim of this study was to determine the accuracies of implant-level impressions for internal-connection dental implants with various angles of divergence. Materials and Methods: Five groups were created according to the angles of divergence (0, 5, 10, 15, and 20 degrees) between two implant analogs in a master model. A master stone model for each group was fabricated. The divergent angle in each master model was verified with a profilometer. Study models (n = 10) from each master model were constructed through the implant-level impression technique. The divergent angle in each study model was verified with the profilometer. The splinted metal framework with passive fit was cast with base metal and cast-on abutments in each master model. The metal framework was tried in the study models and the fit of each framework was evaluated. One-way analysis of variance and the Tukey Honestly Significant Difference test were used for statistical analysis. Results: Statistically significant deformations of the impressions were found in the study models with divergently placed implant analogs. A larger divergence angle resulted in a larger gap between the framework and the implant analogs. The gap in the 20-degree divergence group differed significantly from that measured in the master model (P .01). Conclusions: Internal-connection implants were accurate when the divergence angle was less than 15 degrees. Inaccuracy of impressions increased as the divergence angle increased. The inaccuracy was significantly greater for the 20-degree-divergent implants than the other groups. Palabras clave: deformation of impression, divergent angle, implant impression, implant-level impressions, impression accuracy, internal-connection implant