Three statistical jet noise prediction models are compared for a representative set of single-stream jet cases, which include cold and hot jets of the Strategic Investment in Low-carbon Engine Technology (SILOET) experiment at acoustic Mach number 0.875 and the cold jets of the NASA Small Hot Jet Acoustic Rig (SHJAR) experiment at acoustic Mach numbers 0.5 and 0.9. The implemented models are those proposed by Tam and Auriault, Khavaran, and the Goldstein Generalised Acoustic Analogy (GAA). By the virtue of reduced-order modelling, which is based on the single-point mean flow and turbulence statistics, all these implementations use a number of empirical dimensionless source parameters for far-field noise spectra predictions. In comparison with the Tam and Auriault model, the Khavaran and GAA model implementations use several dimensionless parameters, which are available from the previous literature and assumed to be more-or-less universal for a class of single-stream jets. These parameters include the fluctuating enthalpy function and the dimensionless amplitudes of auto-covariances of turbulent fluctuating stresses and velocities available from the literature. The comparison of the three models is aimed not only at assessing their accuracy for a range of jet conditions, observer angles, and frequencies, but also to examine their robustness outside of a reference jet experiment for which their source models were calibrated. For the input to each model, the mean flow, turbulence kinetic energy, and dissipation rate extracted from Large Eddy Simulations (LES) and Reynolds-Averaged Navier-Stokes (RANS) solutions are considered. Insights and perspectives on using these models in future multidisciplinary design optimisation studies are discussed.