Public transport assignment models have increased in complexity in order to describe passengers' route choices as detailed and correctly as possible. Important trends in the development are (1) timetable-based assignment, (2) inclusion of feeder modes, (3) use of stochastic components to describe differences in passengers' preferences within and between purposes and classes (random coefficients), as well as to describe non-explained variation within a utility theory framework, and (4) consideration of capacity problems at coach level, system level and terminal level. In the Copenhagen-Ringsted Model (CRM), such a large-scale transit assignment model was developed and estimated. The Stochastic User Equilibrium problem was solved by the Method of Successive Averages (MSA). However, the model suffered from very large calculation times. The paper focuses on how to optimise transit assignment models based on MSA combined with a generalised utility function. Comparable tests are carried out on a large-scale network. The conclusion is that there is potential of optimising MSA-based methods. Examples of different approaches for this is presented, tested and discussed in the paper.