Prostheses with single radius (SR) design were supposed to be as good as the physiological kinematic and stability of the knee. This in-vitro biomechanical study compared SR to a multiple radius (MR) design on the one hand and seven left human knee specimens were used. The SR and MR knee prosthesis where implanted with a navigation system. We measured varus/valgus deviation of the mechanical axis and the deviation of the joint-line to the epicondyle-line in different knee flexion degrees (0°, 30°, 45°, 60° and 90°) with and without 15 Nm of varus and valgus stress. Without varus/valgus-stress in all three groups (physiological knee, SR and MR prosthesis) the results were located on the varus-site. The variation of the SR was less than the MR, without being significant. Under varus and valgus stress varus/valgus axis deviation constantly grew. From 0-60° no significant deviation between the two prosthesis models was found. At 90° flexion varus/valgus deviation with the SR component was significantly (p ⩽ 0.05) smaller compared to the MR design. This in-vitro study showed that the SR prosthesis is significantly more stable in the coronal plane than the MR in higher flexion degrees. This could have an improved effect on biomechanical stability with a higher clinical function after SR-TKA.