The use of mixed ionic liquids (ILs) as possible alternatives for the removal of aromatics has been studied here via experimentation and predictions. For the IL mixture, a predictive approach such as COSMO (COnductor like Screening MOdel) along with its extension to RS (real solvent) is used in this work. Initially the COSMO-RS model was benchmarked on three different LLE (liquid-liquid equilibria) systems. The binary LLE ([C(2)mim] [NTf2] (1-ethyl-3-methylimidazolium bis{(trifluoromethyl)sulfonyl}amide) + [P-66614][NTf2] (trihexyl(tetradecyl)phosphonium bis{(trifluoromethyl)sulfonyl}amide) gave a root-mean-square deviation (rmsd) of 7 %. Thereafter, ternary IL systems [EMIM][ETSO4] + benzene + hexane and [EMIM][MESO4] + benzene + hexane gave a rmsd of 1.68 and 1.86 %, respectively. Further the model was benchmarked on the quaternary IL based system: [BMIM][BF4] + [OMIM][BF4] + toluene + heptane, giving an rmsd of 5.25 %. For experimental benclunarking the system [EMIM][ETSO4] (1) + benzene (2) and hexane (3) was examined, which gave an rmsd equal to 4.13 % for the NMR technique and 6.32% via the cloud point method. Further the LLE of the quaternary system [EMIM][ETSO4] + [EMIM][MESO4] + benzene + hexane at T = 298.15 K was determined, and the average rmsd was around 5.25 % for eight tie lines, which is good considering our method to be a priori. Lesser values of selectivity and distribution ratio were obtained for the mixed IL used in this work as compared to a single IL.