Mu-opioid agonists decrease cocaine self-administration in laboratory studies and cocaine use by many cocaine- and opioid-dependent polydrug abusers. To assess the role of mu-agonist efficacy as a determinant of these effects, this study evaluated cocaine- and food-maintained responding by rhesus monkeys (Macaca mulatta) during chronic treatment with saline or the high-efficacy mu-agonist fentanyl (0.001-0.01 mg/kg/h), the intermediate-efficacy mu-agonist morphine (0.032-0.32 mg/kg/h), or the low-efficacy mu-agonists nalbuphine (0.1-1.0 mg/kg/h) and butorphanol (0.0032-0.032 mg/kg/h). Responding was maintained by cocaine and food under a second order schedule of reinforcement during multiple daily sessions of cocaine and food availability. Saline and each opioid dose were administered continuously for 7 consecutive days during availability of each cocaine dose. All four mu-agonists produced dose-dependent and sustained decreases in cocaine self-administration across a range of cocaine doses (0.0032-0.1 mg/kg/injection). Nalbuphine and butorphanol produced the greatest decreases in cocaine self-administration and the smallest effects on food-maintained responding. Morphine and fentanyl produced smaller decreases in cocaine self-administration, and undesirable effects precluded evaluation of higher fentanyl and morphine doses. Decreases in cocaine self-administration produced by nalbuphine and butorphanol probably did not reflect a general blockade of cocaine's abuse-related effects, because nalbuphine and butorphanol did not block the discriminative stimulus effects of cocaine in monkeys trained to discriminate 0.4 mg/kg cocaine from saline in a food-reinforced drug discrimination procedure. These results suggest that low-efficacy mu-agonists may decrease cocaine self-administration to a greater degree and with fewer undesirable effects than high-efficacy mu-agonists.