Cognitive disturbances in major depressive disorder (MDD) constitute a critical treatment target and hold promise as an early predictor of antidepressant treatment response; yet their clinical relevance is not fully established. Therefore, we here investigate if (1) cognitive performance improves over the course of antidepressant treatment and (2) cognitive performance at baseline is predictive of antidepressant treatment response. In the NeuroPharm study (clinical trial id: NCT02869035), 92 antidepressant-free patients with a moderate to severe depressive episode were assessed with a comprehensive cognitive test battery including both cold (emotion-independent) and hot (emotion-dependent) tasks. Patients were tested before and after 12 weeks of standard antidepressant treatment with escitalopram in flexible doses of 10–20 mg. Performance improved across most cognitive domains over the course of antidepressant treatment. Notably, these improvements were independent of improvement in mood symptoms, emphasizing that cognitive disturbances are a distinct symptom and therefore treatment target in MDD. Results did not suggest that performance on any single cognitive measure at baseline was associated with later clinical response to antidepressant treatment. However, a small cluster of patients (N = 28) with globally disturbed cognition at baseline exhibited poorer clinical response after 8 but not 12 weeks of antidepressant treatment, suggesting that severe cognitive disturbances may delay treatment response. Thus, while pretreatment cognitive performance on individual tests may not be useful as clinical markers of treatment response, profiles capturing performance across different cognitive domains may be useful for stratification of patients with MDD and could be helpful in future intervention trials.