Adaptive value-guided decision-making requires weighing up the costs and benefits of pursuing an available opportunity. Though neurons across frontal cortical-basal ganglia circuits have been repeatedly shown to represent decision-related parameters, it is unclear whether and how this information is coordinated. To address this question, we performed large-scale single-unit recordings simultaneously across 5 medial/orbital frontal and basal ganglia regions as rats decided whether to pursue varying reward payoffs available at different effort costs. Single neurons encoding combinations of decision variables (reward, effort, and choice) were represented within all recorded regions. Coactive cell assemblies, ensembles of neurons that repeatedly coactivated within short time windows (<25 ms), represented the same decision variables despite the members often having diverse individual coding properties. Together, these findings demonstrate a multi-level encoding structure for cost-benefit computations where individual neurons are coordinated into larger assemblies that can represent task variables independently of their constituent components.