12C ion therapy has had growing interest in recent years for its excellent dose conformity. However at therapeutic energies, which can be as high as 400 MeV/u, carbon ions produce secondary fragments. For an incident 400 MeV/u 12C ion beam, ~70% of the beam will undergo fragmentation before the Bragg Peak. The dosimetric and radiobiological impact of these fragments must be accurately characterised, as it can result in increasing the risk of secondary cancer for the patient as well as altering the relative biological effectiveness. This work investigates the accuracy of three different nuclear fragmentation models available in the Monte Carlo Toolkit Geant4, the Binary Intranuclear Cascade (BIC), the Quantum Molecular Dynamics (QMD) and the Liege Intranuclear Cascade (INCL++). The models were benchmarked against experimental data for a pristine 400 MeV/u 12C beam incident upon a water phantom, including fragment yield, angular and energy distribution. For fragment yields the three alternative models agreed between ~5 and ~35% with experimental measurements, the QMD using the “Frag” option gave the best agreement for lighter fragments but had reduced agreement for larger fragments. For angular distributions INCL++ was seen to provide the best agreement among the models for all elements with the exception of Hydrogen, while BIC and QMD was seen to produce broader distributions compared to experiment. BIC and QMD performed similar to one another for kinetic energy distributions while INCL++ suffered from producing lower energy distributions compared to the other models and experiment.