Cognitive flexibility is necessary for efficient disengagement from a previous task or reconfiguring and implementing new responses to a novel demand. The orbitofrontal cortex is associated with value encoding and is crucial for flexibility to switch behaviours in front of unexpected outcomes. However, the neuronal mechanisms behind such flexibility to modify the encoded value are unknown. We aim to determine the dynamical mechanisms of orbitofrontal neuronal circuits involved in value-based decisions across different environmental contingencies. For this purpose, we perform large-scale electrophysiological recordings in head-fixed mice during a value-based decision-making task. This task is based on the decision between two options with probabilistic rewards. The reward probability of one option changes in an unannounced way to modify the context of advantage between both options. By training a nonstationary generative recurrent neural network, we reconstruct the dynamics underlying the neuronal recordings. The model allows us to determine the computational mechanisms underlying task performance from a dynamical systems perspective. Our results indicate the proficient performance of mice in the task, as they adapt their choices to maximise the reward. In the model we are able to reconstruct the temporal and geometrical structure of the neuronal activity in the orbitofrontal cortex, achieving the same prediction accuracy for behavioral events as that obtained from the neurophysiological data. In conclusion, we use a nonstationary recurrent neural network to unravel dynamical mechanisms of the adaptability of the orbitofrontal cortex during a value-based decision-making task.Grant-I-5458 of the Austrian Science Fund and FOR5159 of the German Research Foundation