As new generations of thin-film semiconductors are moving toward solution-based processing, the development of printing formulations will require information pertaining to the free energies of mixing of complex mixtures. From the standpoint of in silico material design, this move necessitates the development of methods that can accurately and quickly evaluate these formulations in order to maximize processing speed and reproducibility. Here, we make use of molecular dynamics (MD) simulations, in combination with the two-phase thermodynamic (2PT) model, to explore the free energy of mixing surfaces for a series of halogenated solvents and high-boiling point solvent additives used in the development of thin-film organic semiconductors. Although the combined methods generally show good agreement with available experimental data, the computational cost to traverse the free-energy landscape is considerable. Hence, we demonstrate how a Bayesian optimization scheme, coupled with the MD and 2PT approaches, can drastically reduce the number of simulations required, in turn shrinking both the computational cost and time.