Impression techniques should precisely represent the 3-dimensional status of implants to allow for the fabrication of passively fitting prostheses and subsequently the elimination of strain on supporting implant components and surrounding bone. The aim of this study was to compare the accuracy of an abutment level impression method with that of an implant level (direct and indirect) impression method using polyether impression material to obtain precise definitive casts and prostheses. A reference acrylic resin dentoform with 2 internal connection implants (Implantium) was made. A total of 21 medium-consistency polyether impressions of the dentoform, including 7 direct implant level, 7 indirect implant level, and 7 abutment level (after 2 straight abutments were secured), were made. Impressions were poured with American Dental Association (ADA) type IV stone, and the positional accuracy of the implant replica heads and abutment analogs in each dimension of x-, y-, and z-axes, as well as angular displacement (Δθ), was evaluated using a coordinate measuring machine. Noble alloy 3-unit castings were fabricated and seated on the abutments in 3 groups; marginal discrepancies were measured at 4 points between prostheses and abutments. Data were analyzed using Mann-Whitney U test, 1-way analysis of variance (ANOVA), and Kruskal-Wallis tests. In comparisons of different impression techniques, only significant statistical Δθ differences were noted between the abutment level method and other techniques (P < .001). Results of this study reveal that although the implant level impression method could better transfer the angular position of the implants (Δθ), the impression method could not affect Δy, Δx, and Δz coordinates of the implants or marginal discrepancy of the 3-unit fixed partial dentures (FPD).