This is the first study to use stochastic frontier analysis to simultaneously estimate the technical, cost and profit efficiency of physician practices for different physician specialist groups. We base our analysis on a unique panel data set of 4964 physician practices in Germany for the years 2008 to 2010. The data contains information on practice costs and revenues, services provided, as well as physician characteristics and practice characteristics. Additionally we consider a wide range additional variables not previously analyzed in this context (e.g. sub-specialization of physician groups and environmental factors such as physician density in the area). We investigate differences in cost, technical and profit efficiency utilizing production−/cost- and profit-functions with a translog functional form. We estimated the stochastic frontier using the comprehensive one-step approach for panel data following Battese and Coelli (Empir Econ 20(2): 325–332, 10). Overall findings indicate that participation in disease management programs and the degree of specialization are associated with significantly higher technical- cost-, and profit-efficiency for most physician specialist groups, e.g. due to the standardization of processes. In addition, our analyses show that group practices perform significantly better than single practices. This may be due to indivisibilities in expensive technical equipment, which can lead to different health care services being provided by different practice types. A more thorough look at specialist groups suggests that it is important to investigate all efficiency types for different physician groups, as results may depend on the type of efficiency analyzed as well as the physician group in question.