Isolating specific cognitive effects of antidepressant drugs is crucial to develop targeted and individualized treatment selection in psychiatry. In this double-blind, placebo-controlled study in healthy controls, we used computational modeling to characterize the cognitive effects of two classes of drugs for depression, escitalopram, a typical SSRI which increases serotonergic transmission, and agomelatine, which activates melatonin receptors and antagonizes 5-HT2C serotonergic receptors. 128 healthy participants were randomized to receive either escitalopram (20 mg), agomelatine (25 mg or 50 mg) or placebo for 8 weeks and performed two complementary learning tasks at three time-points allowing to measure early (3 days), intermediate (2 weeks) and delayed (8 weeks) treatment effects. The first task was a simple probabilistic instrumental learning task evaluating how participants learned from positive and negative feedback. The second task was a more complex reversal learning task devised to assess learning from positive and negative feedback in an unstable environment. At 8 weeks, both drugs improved accuracy in task 1 and decreased choice stochasticity in task 2 compared to placebo. Agomelatine 25 and 50 mg had an additional early beneficial effect on reward processing at 3 days whereas agomelatine 50 mg showed maximal effects at 2 weeks. Our study provides one of the very first cognitive evaluations of the delayed effects of antidepressant drugs in healthy volunteers. It reveals that they share common beneficial effect on learning along with pharmacological-specific effects. All observed effects varied highly over time, highlighting the non-linearity of the cognitive impact.