In this paper, we use an unprecedentedly large sample (158) of confirmed strong lens systems for model selection, comparing five well-studied Friedmann-Robertson-Walker cosmologies: {Lambda}CDM, wCDM (the standard model with a variable dark-energy equation of state), the R_h_=ct universe, the (empty) Milne cosmology, and the classical Einstein-de Sitter (matter-dominated) universe. We first use these sources to optimize the parameters in the standard model and show that they are consistent with Planck, though the quality of the best fit is not satisfactory. We demonstrate that this is likely due to underreported errors, or to errors yet to be included in this kind of analysis. We suggest that the missing dispersion may be due to scatter about a pure single isothermal sphere (SIS) model that is often assumed for the mass distribution in these lenses. We then use the Bayes information criterion, with the inclusion of a suggested SIS dispersion, to calculate the relative likelihoods and ranking of these models, showing that Milne and Einstein-de Sitter are completely ruled out, while R_h_=ct is preferred over {Lambda}CDM/wCDM with a relative probability of ~73 percent versus ~24 percent. The recently reported sample of new strong lens candidates by the Dark Energy Survey, if confirmed, may be able to demonstrate which of these two models is favoured over the other at a level exceeding 3{sigma}.