MELCOR has become the preferred code for severe accident analysis in Switzerland, on account of its integrated systems-level approach and validation against experiments and more detailed codes. The present analysis of the TMI-2 accident is part of a programme to assess independently this approach. The MELCOR 1.8.5RD calculations were based on a model for phases 1 and 2 provided by the code developers. This was extended to accept boundary conditions for phases 3 and 4, while simplifying the thermal hydraulic noding in certain regions. These calculations, since extended to cover radionuclide transport and containment, are the first performed with MELCOR for the accident sequence, through to the end of phase 4. A similar approach was taken with SCDAPSIM/MOD3.4 and SCDAP/RELAP5/MOD3.2. Uncertainties in the actual boundary conditions and plant data make this a challenging task. Comparisons between the calculated results and observed data such as primary system pressure, hot leg temperatures and core liquid level show fair to good agreement, bearing in mind the above uncertainties. The codes were able to reproduce the major safety-related accident signatures for suitable chosen values for degradation parameters. The results provide a good basis for the NPP analysis foreseen, with the proviso that limitations in the modelling of core degradation processes imply use of sensitivity studies. These limitations reflect the current state of knowledge.