Aims. Bi-unicondylar arthroplasty (Bi-UKA) is a bone and anterior cruciate ligament (ACL)-preserving alternative to total knee arthroplasty (TKA) when the patellofemoral joint is preserved. The aim of this study is to investigate the clinical outcomes and biomechanics of Bi-UKA. Methods. Bi-UKA subjects (n = 22) were measured on an instrumented treadmill, using standard gait metrics, at top walking speeds. Age-, sex-, and BMI-matched healthy (n = 24) and primary TKA (n = 22) subjects formed control groups. TKA subjects with preoperative patellofemoral or tricompartmental arthritis or ACL dysfunction were excluded. The Oxford Knee Score (OKS) and EuroQol five-dimension questionnaire (EQ-5D) were compared. Bi-UKA, then TKA, were performed on eight fresh frozen cadaveric knees, to investigate knee extensor efficiency under controlled laboratory conditions, using a repeated measures study design. Results. Bi-UKA walked 20% faster than TKA (Bi-UKA mean top walking speed 6.7 km/h (SD 0.9),TKA 5.6 km/h (SD 0.7), p < 0.001), exhibiting nearer-normal vertical ground reaction forces in maximum weight acceptance and mid-stance, with longer step and stride lengths compared to TKA (p < 0.048). Bi-UKA subjects reported higher OKS (p = 0.004) and EQ-5D (p < 0.001). In vitro, Bi-UKA generated the same extensor moment as native knees at low flexion angles, while reduced extensor moment was measured following TKA (p < 0.003). Conversely, at higher flexion angles, the extensor moment of TKA was normal. Over the full range, the extensor mechanism was more efficient following Bi-UKA than TKA (p < 0.028). Conclusion. Bi-UKA had more normal gait characteristics and improved patient-reported outcomes, compared to matched TKA subjects. This can, in part, be explained by differences in extensor efficiency. Cite this article: Bone Joint Res 2021;10(11):723–733