One goal of the QUELL experiment is the validation of numerical codes for the simulation of thermo-hydraulic transients in Cable-in-Conduit Conductors with Central Channel (C-5). The validation of Gandalf has focused on 17 quench propagation runs. In general, the good global agreement of the simulations with the experiments-the standard deviation is in the range between 5% and 25%-as well as the good computer performance qualifies Gandalf as a reliable engineering tool. One limitation of the simplified two-flow model implemented in Gandalf is that it tends to overestimate the quench propagation, and hence the helium pressure increase, in the initial phase of the quench evolution, and to underestimate them in the final phase. In standard quenches this limitation can be overcome by full two-flow modeling (Mithrandir code). Two further problem areas have been identified: the high sensitivity (in common with full two-flow modeling) to conductor parameters, if the coil is operated at low temperature margin, and the lack of accuracy when using simplified hydraulic boundary conditions. (C) 1998 Elsevier Science Ltd.