Canine cranial cruciate ligament rupture is often bilateral and asymmetrical, ranging from partial to complete rupture. The purpose of our diagnostic accuracy study was to assess the accuracy of 3 Tesla magnetic resonance imaging (MRI) detection of fiber loss and use of a visual analog scale in the diagnosis of complete versus partial cranial cruciate ligament rupture in 28 clinical dogs with unilateral complete rupture and contralateral partial rupture. Three Tesla MRI was performed on 56 stifles using sagittal sequences (T2-weighted fast spin echo with fat saturation, proton density fast spin echo, and T2-weighted 3D fast spin echo CUBE). Two MRI observers assessed the cranial cruciate ligament for fiber loss and completed a visual analog scale. The MRI data were compared to arthroscopy and clinical status. Accuracy classifying partial or complete rupture was assessed using receiver operating characteristic analysis. Compared to arthroscopy, for complete cranial cruciate ligament rupture, sensitivity, specificity, and accuracy of MRI detection of fiber loss were 0.78, 0.50-0.60, and 0.68-0.71, respectively, and, for partial tears, specificity was 1.00. An MRI In stifles with clinically unstable complete cranial cruciate ligament rupture, both MRI tests are sensitive though not specific compared to arthroscopy. As a diagnostic imaging method, MRI may help guide treatment in patients with cranial cruciate ligament damage, particularly for stable partial rupture.