We aim to study technical efficiency of undergraduate teaching activity in national (public) universities in Argentina. We employ a stochastic frontier analysis (SFA) with a panel of 37 national universities over 2005-2013. We compare models that do not account for heterogeneity with heterogeneity-extended SFA models. We find from 18 percent to 25 percent of inefficiency on average in terms of lost outcomes (graduates) depending on the specification, with high dispersion among universities. Models considering heterogeneity report the highest levels of technical efficiency. Besides, the results show evidence of heterogeneity both observed and unobserved. Our estimates are robust to different specifications.