The literature contains limited yet controversial information regarding whether a fixed or a mobile bearing implant should be used in unicompartmental knee arthroplasty (UKA). This randomized study was to further document the performance and comparison of the two designs. Fifty-six knees in 48 patients (mean age of 72 years) undergoing medial UKA were randomized into a fixed bearing (Miller/Galante) or a mobile bearing (Oxford) UKA. The 2 year clinical outcomes (clinical scores), radiographic findings, and weight bearing knee kinematics (assessed using RSA) were compared between the two groups. The mobile bearing knees displayed a larger and an incrementally increased tibial internal rotation (4.3 degrees, 7.6 degrees, 9.5 degrees vs. 3.0 degrees, 3.0 degrees, 4.2 degrees respectively at 30 degrees, 60 degrees, 90 degrees of knee flexion) compared to the fixed ones. The medial femoral condyle in the mobile bearing knees remained 2 mm from the initial position vs. a 4.2 mm anterior translation in the fixed bearing knees during knee flexion. The contact point in the mobile bearing implant moved 2 mm posteriorly vs. a 6 mm anterior movement in the other group. The mobile bearing knees had a lower incidence of radiolucency at the bone implant interface (8% vs. 37%, p < 0.05). The incidence of lateral compartment OA and progression of OA at patello-femoral joint were equal. No differences were found regarding Knee Society Scores, WOMAC, and SF-36 scores (p > 0.05). This study indicates that mobile bearing knees had a better kinematics, a lower incidence of radiolucency but not yet a better knee function at 2 years.