Acinetobacter baumannii is among the most dangerous pathogens and emergence of resistance is highly problematic. Our objective was to identify and rationally optimize ?-lactam-plus-aminoglycoside combinations via novel mechanism-based modeling that synergistically kill and prevent resistance of carbapenem-resistant A. baumannii. We studied combinations of 10 ?-lactams and three aminoglycosides against four A. baumannii strains, including two imipenem-intermediate (MIC, 4 mg/liter) and one imipenem-resistant (MIC, 32 mg/liter) clinical isolate, using high-inoculum static-concentration time-kill studies. We present the first application of mechanism-based modeling for killing and resistance of A. baumannii using Monte Carlo simulations of human pharmacokinetics to rationally optimize combination dosage regimens for immunocompromised, critically ill patients. All monotherapies achieved limited killing (5 log10) and prevented regrowth of all four strains. Modeling demonstrated that imipenem likely killed the aminoglycoside-resistant population and vice versa and that aminoglycosides enhanced the target site penetration of imipenem. Against carbapenem-resistant A. baumannii (MIC, 32 mg/liter), optimized combination regimens (imipenem at 4 g/day as a continuous infusion plus tobramycin at 7 mg/kg of body weight every 24 h) were predicted to achieve >5 log10 killing without regrowth in 98.2 of patients. Bacterial killing and suppression of regrowth were best achieved for combination regimens with unbound imipenem steady-state concentrations of at least 8 mg/liter. Imipenem-plus-aminoglycoside combination regimens are highly promising and warrant further evaluation.