Self-administration procedures have not yet provided evidence that freely moving mice can reliably acquire and maintain an operant behavior to self-administer cannabinoid agonists. The aim of the present work was to establish a model of cannabinoid operant intravenous self-administration in freely moving mice given the relevance of this species for the use of genetically modified animals. In addition, the possible involvement of the k/dynorphin system in cannabinoid self-administration was evaluated by using pro-dynorphin knockout mice. Outbred CD1 wild-type mice as well as pro-dynorphin knockout and wild-type mice were trained to self-administer the cannabinoid receptor agonist WIN 55,212-2 under an FR1 schedule of reinforcement. Two cannabinoid training doses (6.25 and 12.5 mg/ kg/infusion) were used in the acquisition studies in outbred mice. Animals acquired a reliable operant responding to self-administer WIN 55,212-2 (12.5 mg/kg/infusion), but required as many as 15 sessions to attain this behavior. Interestingly, when a previous injection of WIN 55,212-2 (0.1 mg/kg, i.p.) was administered in the home-cage 24 h before the first session, mice acquired operant responding for cannabinoid self-administration by the fourth session. When the k-opioid agonist antagonist nor-binaltorphimine (5 mg/kg s.c.) was administered 4 h before the first session, the time required to acquire a reliable cannabinoid self-administration was also significantly reduced. Finally, a shift to the left in the dose-intake curve to self-administer WIN 55,212-2 was observed in pro-dynorphin knockout mice when compared to wild-type mice. These results indicate that the activation of the k/dynorphin opioid system after WIN 55,212-2 administration could counteract cannabinoid rewarding effects.