PURPOSE. The aim of this study was to evaluate the interface accuracy of computer-assisted designed and manufactured (CAD/CAM) titanium abutments and implant fixture compared to gold-cast UCLA abutments. MATERIALS AND METHODS. An external connection implant system (Mark III, n=10) and an internal connection implant system (Replace Select, n=10) were used, 5 of each group were connected to milled titanium abutment and the rest were connected to the gold-cast UCLA abutments. The implant fixture and abutment were tightened to torque of 35 Ncm using a digital torque gauge, and initial detorque values were measured 10 minutes after tightening. To mimic the mastication, a cyclic loading was applied at 14 Hz for one million cycles, with the stress amplitude range being within 0 N to 100 N. After the cyclic loading, detorque values were measured again. The fixture-abutment gaps were measured under a microscope and recorded with an accuracy of ${pm}0.1{mu}m$ at 50 points. RESULTS. Initial detorque values of milled abutment were significantly higher than those of cast abutment (P<.05). Detorque values after one million dynamic cyclic loadings were not significantly different (P>.05). After cyclic loading, detorque values of cast abutment increased, but those of milled abutment decreased (P<.05). There was no significant difference of gap dimension between the milled abutment group and the cast abutment group after cyclic loading. CONCLUSION. In conclusion, CAD/CAM milled titanium abutment can be fabricated with sufficient accuracy to permit screw joint stability between abutment and fixture comparable to that of the traditional gold cast UCLA abutment.