Rationale: 16p11.2 duplication is associated with numerous neuropsychiatric conditions at a genome-wide level, including psychosis. Mice modeling 16p11.2 duplication may provide important insights into cognitive risk factors, in particular in reward-guided decision making. NMDAR function has also been implicated in psychosis phenotypes, but whether these phenotypes differ by genetic risk factor is unknown. Objectives: We aimed to: 1) identify sex and genotype differences in early operant training and two-arm spatial restless bandit task performance; 2) examine the effects of an NMDAR antagonist on task performance and strategy across genotypes. Methods: 16p11.2 duplication and wildtype mice completed a series of training schedules of escalating difficulty followed by bandit tasks. MK-801 and saline were administered in alternating sessions prior to later bandit task performance. Results: Large sex differences in early operant training revealed some male-biased impacts of 16p11.2 duplication, contingent on training schedule difficulty. Once on the two-arm spatial restless bandit task, 16p11.2 duplication was no longer a strong contributor to decision making. However, MK-801 decreased the tendency to stay with a rewarded choice, lowered the probability of selecting the highest rewarded option, and decreased the influence of prior outcomes on choice. Conclusions: The male-biased vulnerability in early operant training suggests that strategies for learning early schemas or in novel environments may be impacted by 16p11.2 duplication in males. In contrast, NMDAR are influential in the ability to flexibly switch between choices, and disrupting this function significantly impairs decision making in all animals.