A substantial number of public water system wells in south-central Texas withdraw groundwater from the karstic, highly productive Edwards aquifer. However, the use of numerical groundwater flow models to aid in the delineation of contributing areas for public water system wells in the Edwards aquifer is problematic because of the complex hydrogeologic framework and the presence of conduit-dominated flow paths in the aquifer. The U.S. Geological Survey, in cooperation with the Texas Commission on Environmental Quality, evaluated six published numerical groundwater flow models (all deterministic) that have been developed for the Edwards aquifer San Antonio segment or Barton Springs segment, or both. This report describes the models developed and evaluates each with respect to accessibility and ease of use, range of conditions simulated, accuracy of simulations, agreement with dye-tracer tests, and limitations of the models. These models are (1) GWSIM model of the San Antonio segment, a FORTRAN computer-model code that pre-dates the development of MODFLOW; (2) MODFLOW conduit-flow model of San Antonio and Barton Springs segments; (3) MODFLOW diffuse-flow model of San Antonio and Barton Springs segments; (4) MODFLOW Groundwater Availability Modeling [GAM] model of the Barton Springs segment; (5) MODFLOW recalibrated GAM model of the Barton Springs segment; and (6) MODFLOW–DCM (dual conductivity model) conduit model of the Barton Springs segment. The GWSIM model code is not commercially available, is limited in its application to the San Antonio segment of the Edwards aquifer, and lacks the ability of MODFLOW to easily incorporate newly developed processes and packages to better simulate hydrologic processes. MODFLOW is a widely used and tested code for numerical modeling of groundwater flow, is well documented, and is in the public domain. These attributes make MODFLOW a preferred code with regard to accessibility and ease of use. The MODFLOW conduit-flow model incorporates improvements over previous models by using (1) a user-friendly interface, (2) updated computer codes (MODFLOW–96 and MODFLOW–2000), (3) a finer grid resolution, (4) less-restrictive boundary conditions, (5) an improved discretization of hydraulic conductivity, (6) more accurate estimates of pumping stresses, (7) a long transient simulation period (54 years, 1947–2000), and (8) a refined representation of high-permeability zones or conduits. All of the models except the MODFLOW–DCM conduit model have limitations resulting from the use of Darcy’s law to simulate groundwater flow in a karst aquifer system where non-Darcian, turbulent flow might actually dominate. The MODFLOW–DCM conduit model is an improvement in the ability to simulate karst-like flow conditions in conjunction with porous-media-type matrix flow. However, the MODFLOW–DCM conduit model has had limited application and testing and currently (2008) lacks commercially available pre- and post-processors. The MODFLOW conduit-flow and diffuse-flow Edwards aquifer models are limited by the lack of calibration for the northern part of the Barton Springs segment (Travis County) and their reliance on the use of the calibrated hydraulic conductivity and storativity values from the calibrated Barton Springs segment GAM model. The major limitation of the Barton Springs segment GAM and recalibrated GAM models is that they were calibrated to match measured water levels and springflows for a restrictive range of hydrologic conditions, with each model having different hydraulic conductivity and storativity values appropriate to the hydrologic conditions that were simulated. The need for two different sets of hydraulic conductivity and storativity values increases the uncertainty associated with the accuracy of either set of values, illustrates the non-uniqueness of the model solution, and probably most importantly demonstrates the limitations of using a one-layer model to represent the heterogeneous hydrostratigraphic units composing the Edwards aquifer. In general, the best matches or agreement between groundwater flow directions inferred by numerical model simulation, and by dye-tracer tests, are observed where model outputs accurately reproduce the configuration of the potentiometric surface with regard to the positions of major and minor groundwater troughs and divides. None of the models, with the possible exception of the MODFLOW–DCM conduit model, has a documented capability to accurately simulate travel times for conduit-dominated velocities in the Edwards aquifer. Public water system assessments of wells in the Barton Springs segment of the Edwards aquifer, and elsewhere where conduit-flow conditions are thought to dominate aquifer hydraulic behavior, might be enhanced by use of either the MODFLOW–DCM model or the newly developed U.S. Geological Survey MODFLOW Conduit-Flow Process module for MODFLOW–2005, because each incorporates a type of dual or triple hydraulic conductivity approach and has the capability to explicitly simulate turbulent flow and conduit hydraulic characteristics.