The use of 1-methyl-2-pyrrolidone (NMP) as an entrainer in extractive distillation (ED) has raised safety and environmental concerns. This study investigated biobased and greener solvents as sustainable entrainers for the industrially relevant separations of an azeotropic n-hexane-ethanol (HEX-ETH) mixture and a close-boiling methylcyclohexane-toluene (MCH-TOL) mixture. A multi-stage entrainer selection framework integrating computational predictions, experimental validation, and process-level assessment was applied to identify effective greener entrainers. A comprehensive assessment for the economic and sustainability performances was conducted, including total annual cost (TAC), energy intensity, CO₂ emissions, and water consumption. Sensitivity analyses were further performed to evaluate the effect of recycled entrainer purity, payback period, and entrainer price on the TAC. The results indicate that the investigated greener entrainers generally achieve economic and sustainability performances comparable to those of NMP while offering substantially lower toxicity. Considering the estimated TAC uncertainty of approximately below 11%, greener entrainer-based ED processes exhibit TAC differences within the estimation uncertainty of the corresponding NMP-based processes. However, HEX-ETH-dimethyl isosorbide (DMI) and MCH-TOL-1-butylpyrrolidin-2-one (NBP), which show TAC increases of up to 14.2 and 15.1%, respectively, are mainly due to higher CAPEX for HEX-ETH-DMI and higher OPEX for MCH-TOL-NBP. Overall, the greener entrainers provide comparable economic and favorable sustainability performance relative to NMP, with an energy intensity ranging from 0.31 to 0.38 kWh/kg_products, CO₂ emissions decrease by 5.8% for MCH-TOL-gamma-valerolactone (GVL), while slightly higher 1.5–20.0% are observed in the other cases. Water consumption is similar for most systems, except for HEX-ETH-DMI and MCH-TOL-NBP, which show 25.0% higher values. Sensitivity analysis confirms that entrainer ranking remains robust under variations in entrainer purity constraint and payback periods, whereas entrainer price significantly affects economic competitiveness. Supporting analyses showed that selectivity exhibits better descriptive consistency with TAC trends than capacity, especially for MCH-TOL, whereas both metrics shows weaker descriptive consistency with TAC trends for the associating HEX-ETH mixture. Predictive models of UNIFAC and COSMO-SAC yielded TAC deviations below 31% and 32%, respectively, when compared with NRTL-based simulations. This indicates their suitability for early-stage entrainer screening for systems without strong association or that do not exhibit abnormal phase behavior. However, the COSMO-SAC prediction for the ethanol-DMI mixture produces nonphysical azeotropic behavior and leads to process convergence failure. This highlights the risk of applying COSMO-SAC to strongly associating mixtures without prior VLE validation. Therefore, validation with experimental data is required before applying predictive models in process-level simulations. ... Read more