Based on knowledge obtained from experimental programs, considerable progress has been made in severe accident code development. The three leading severe accident codes used in the U.S., MELCOR, MAAP4 and SCDAP/RELAP5, are compared herein as part of an evaluation of the relative state of severe accident modeling in each of the codes. The MELCOR code is evolving from a probabilistic risk assessment tool to a best-estimate severe accident system analysis code. Some advantages that MELCOR has are the capability to evaluate containment behavior and the source term to the environment, and the great modeling flexibility that the control volume approach and control functions afford. The MAAP4 code was developed to perform fast-running full simulations of severe accidents. Due in part to the simplified form of the conservation equations and the coarser discretization of the reactor systems, MAAP4 has calculation times far shorter than those of the other codes while producing credible results. The SCDAP/RELAP5 code contains more mechanistic physics models than the other codes for both severe accident and thermal-hydraulic phenomena and has undergone extensive validation against plant and experimental data. The codes' overall attributes, relevant physics models and calculation results are compared herein. A hypothetical TMLB' scenario (station blackout with no recovery of auxiliary feedwater) at a 4-loop pressurized water reactor (PWR) was calculated by all three codes. Detailed plots showed that, despite considerable differences in the codes themselves, the calculation results of the codes are very similar in terms of thermal-hydraulic and core degradation response. There are minor discrepancies in various timings of phenomena, which are within the uncertainties of the code numerical computation and the physics models. The thermal challenge to the primary loop radionuclide barrier, such as the steam generator tube heat structures, is also compared among the three codes.