Diffusion-weighted MR imaging (DWI) is an essential tool in oncology. Diffusion models beyond monoexponential fitting attempt to capture non-Gaussian decay using additional data acquisition. Model repeatability and suitability is critical, but often neglected. We report findings from fitting multiple diffusion models in a benign breast cancer repeatability cohort, and show no clear dominance of diffusion models across voxels, patients, or scans. Repeatability of ADC, IVIM, and stretched exponential parameters are reported, and highlight the complexity of making inferences from DWI parameters.The potential of DWI in oncology is tempered by a need for critical appraisal of the model and parameter applicability.