STATEMENT OF PROBLEM: Currently, many implant systems are developed and divided into two types according to their joint connection: external or internal connection. Regardless of the connection type, screw loosening is the biggest problem in implant-supported restoration. PURPOSE: The purpose of this study is to assess the difference in stability of abutment screws between the external and internal hexagonal connection types under cyclic loading. MATERIAL AND METHODS: Each of the 15 samples of external implants and internal abutments were tightened to 30 N/cm with a digital torque gauge, and cemented with a hemispherical metal cap. Each unit was then mounted in a $30^{circ}$ inclined jig. Then each group was divided into 2 sub-groups based on different periods of cyclic loading with the loading machine (30 N/ cm - 300 N/cm,14 Hz: first group $1{ imes}10^6$, $5{ imes}10^6$ cyclic loading; second group $3{ imes}10^6$, $3{ imes}10^6$ for a total cyclic loading of $6{ imes}10^6$) The removal torque value of the screw before and after cyclic loading was checked. SPSS statistical software for Windows was used for statistical analysis. Group means were calculated and compared by ANOVA, independent t-test, and paired t-test with ${alpha}$=0.05. RESULTS: In the external hexagonal connection, the difference between the removal torque value of the abutment screw before loading, the value after $1{ ims}10^6$ cyclic loading, and the value after $1{ imes}10^6$, and additional $5{ imes}10^6$ cyclic loading was not significant. The difference between the removal torque value after $3{ imes}10^6$ cyclic loading and after $3{ imes}10^6$, and additional $3{ imes}10^6$ cyclic loading was not significant. In the internal hexagonal connection, the difference between the removal torque value before loading and the value after $1{ imes}10^6$ cyclic loading was not significant, but the value after $1{ imes}10^6$, and additional $5{ imes}10^6$ cyclic loading was reduced and the difference was significant (P < .05). In addition, in the internal hexagonal connection, the difference between the removal torque value after $3{ imes}10^6$ cyclic loading and the value after $3{ imes}10^6$, and additional $3{ imes}10^6$ cyclic loading was not significant. CONCLUSION: The external hexagonal connection was more stable than the internal hexagonal connection after $1{ imes}10^6$, and additional $5{ imes}10^6$ cyclic loading (t = 10.834, P < .001). There was no significant difference between the two systems after $3{ imes}10^6$, and additional $3{ imes}10^6$ cycles.