Building and deploying data-derived models has become a ubiquitous activity in many fields. Generation of vastly complex models is now possible where the limiting factor within many fields of application is a sufficiently large and diverse dataset with which one can train models. For many tasks, financial or time costs limit the collection of data at the desired accuracy, and so methods which are able to take advantage of multiple disparate sources of data are beginning to become popular. One example of this is the emerging area of multifidelity optimization Song et al. (2018); Huang et al. (2006); Kandasamy et al. (2017) where optimisation algorithms are able to make use of queries of approximate variants or lower'fidelities' of the intended optimisation target. For single fidelity blackbox optimisation problems there are well known benchmark suites Hansen et al. (2021) that allow comparison of different algorithms. However preparation of such benchmarks for multifidelity optimisation is challenging due to the need not only to specify diverse and relevant optimisation problems but also multiple different proxies. Recent libraries of analytic functions suitable for multifidelity optimisation have been developed van Rijn and Schmitt (2020); Mainini et al. (2022) as have some general benchmark suites Wang et al. (2018); Eggensperger et al. (2021). However the lower fidelity approximations of current benchmarks do not offer fine grained tools for controlling the behaviour of the low fidelity proxies.