There is limited validation of MONK using 232Th/ 233U benchmarks, and existing validation of Monte Carlo codes has generally used older nuclear data libraries. Here we use MONK-9A to systematically study a series of thermal 232Th/ 233U-based benchmark experiments, taken from the International Criticality Safety Benchmark Evaluation Project handbook and using JEF-2.2 and JEFF-3.1 libraries. We find that 232Th-dominated thermal systems (those moderated with graphite or light water) with limited 233U content show little difference when modelled with JEF-2.2 or JEFF-3.1. However, when there is significant 233U then JEFF-3.1 is found to give considerably better agreement with experiment. There is likely to be little benefit in utilising JEFF-3.1 instead of JEF-2.2 when modelling systems for which the predominant nuclides are 232Th, 235U and 238U. However, for systems containing large fractions of 233U – such as light water breeder reactors operating on a 232Th-233U cycle – it is recommended that JEFF-3.1 is utilised.