There is a growing demand in modeling and simulation to capture and predict increasingly intricate details of complex physical systems including fluid dynamics, structural mechanics, or multiphysics phenomena. This growing demand, coupled with the use of fine spatial and temporal resolutions, generally leads to the development of large-scale high-fidelity models that exhibit superior predictive capability of a system’s behavior. However, the increased fidelity comes with significant trade-offs, including higher computational demands and data storage limitations. These challenges are further aggravated in outer-loop applications, such as optimization, design, or uncertainty quantification, where multiple model evaluations are required. A promising approach to mitigate these challenges is the development of lower fidelity surrogate models which approximate the high-fidelity model and incur a lower computational cost. The focus of this minisymposium is to showcase novel surrogate modelling methods that accelerate large-scale high-fidelity simulations, and that address the critical challenges associated with their effective use in real-world applications.