The paper examines Kant’s views on divine foreknowledge of contingent truths, in particular truths concerning free actions of creatures. It first considers the shape this traditional philosophical problem takes in the transcendental idealist context. It then situates Kant’s views relative to three competing theories of foreknowledge discussed by Leibniz. These are Molina’s theory of middle knowledge, the Thomist theory of foreknowledge through divine predeterminations, and Leibniz’s own ‘possible worlds’ theory. The paper concludes that no consistent theory of divine foreknowledge emerges in Kant’s philosophy. His discussions alternate between two inadequate and incompatible models. One is a post-volitional model suggested by his conception of the intuitive intellect’s relation to creation. Extended to creaturely free action, it is incompatible with his commitment to the relative autonomy of free action. The other is a version of Leibniz’s possible worlds solution; it cannot underwrite certain foreknowledge of determinate outcomes in a libertarian setting.