Acoustic travel-time tomography in the atmosphere is based on travel-time measurements of sound signals propagating along different known ray paths through a medium. Because the speed of sound mainly depends on temperature and flow properties, an inversion of these travel times allows an estimation of temperature and wind velocity fields. The main reconstruction techniques for solving such inverse problems are least-squares methods and stochastic inversion algorithms. In this study, five representatives belonging to these types of inverse approaches are evaluated by reconstructions of two-dimensional temperature distributions from synthetically generated and experimental data. The comparison of the reconstruction results reveals several differences between the algorithms concerning spatial resolution of the reconstructed image, accuracy, and computational efficiency. The stochastic approach provides accurate reconstructions of spatially highly resolved temperature fields when the turbulence characteristic is chosen carefully. Nevertheless, the choice of suitable turbulence parameters, the determination of measurement errors prior to an experiment as well as comparatively high memory requirements of this method are unfavorable for real-time analysis of measured data although possible. In contrast, fast and simple on-site interpretations of temperature fields with acceptable accuracy are feasible with least-squares methods.