This paper examines the efficiency of international airlines for the period 1998-2012 using some competing stochastic frontier (SF) panel data models. It estimates a cost function for multi-output airline services, separating passenger and goods transportation at the national and international levels. Our preferred SF model distinguishes airlines heterogeneity from time-invariant persistent inefficiency, and transient (time-varying) inefficiency from random the noise component. This four component SF model is compared with two other competing SF models in which one of the four components is missing. All the models are estimated using the maximum likelihood method. The models give predicted values of persistent, transient and overall efficiency for each airline and time period. The mean and dispersion of cost efficiency amongst airlines differ by model specifications and according to their geographical area of operations. The performance difference may be a consequence of different market structures and deregulation processes and of specific competitive conditions such as resource availability and strategic alliances with competitors. The results confirm that, in general, none of the airlines is able to achieve full cost efficiency. We find that carriers based in the Asia region are more efficient than those operating in the European and North American regions. The bigger airlines are unable to take full advantage of economies of scale and are not more efficient than their smaller cousins.