High-value primary care for high-needs patients-those with multiple physical, mental, or behavioral health conditions-is critical to improving health system performance. However, little is known about what types of physician practices perform best for high-needs patients. We examined two scale-related characteristics that could predict how well physician practices delivered care to this population: the proportion of patients in the practice that were high-needs and practice size (number of physicians). Using four years of data on commercially insured, high-needs patients in Michigan primary care practices, we found lower spending and utilization among practices with a higher proportion of high-needs patients (more than 10 percent of the practice's panel) compared to practices with smaller proportions. Small practices (those with one or two physicians) had lower overall spending, but not less utilization, compared to large practices. However, practices with a substantial proportion of high-needs patients, as well as small practices, performed slightly worse on a composite measure of process quality than their associated reference group. Practices that treat a high proportion of high-needs patients might have structural advantages or have developed specialized approaches to serve this population. If so, this raises questions about how best to make use of this knowledge to foster high-value care for high-needs patients.