Neurocognitive deficits in schizophrenia can reach 1 to 2 standard deviations below healthy controls. The comparative effect of typical and atypical antipsychotic medications on neurocognition is controversial, and based primarily on studies with small samples and large doses of typical comparator medications. The present study assessed neurocognitive efficacy. It was hypothesized that olanzapine treatment would improve neurocognitive deficits to a greater degree than either risperidone or haloperidol treatment. This was a double-blind, randomized, controlled, parallel study with neurocognition assessed at baseline, and 8, 24, and 52 weeks. Per protocol, the haloperidol arm was discontinued. Four hundred and fourteen inpatients or outpatients with schizophrenia and schizoaffective disorder were treated with oral olanzapine (n = 159), risperidone (n = 158), or haloperidol (n = 97). Individual domains (executive function, learning and memory, processing speed, attention/vigilance, verbal working memory, verbal fluency, motor function, and visuospatial ability) were transformed into composite scores and compared between treatment groups. At the 52-week endpoint, neurocognition significantly improved in each group (p < 0.01 for olanzapine and risperidone, p = 0.04 for haloperidol), with no significant differences between groups. Olanzapine- and risperidone-treated patients significantly (p < 0.05) improved on domains of executive function, learning/memory, processing speed, attention/vigilance, verbal working memory, and motor functions. Additionally, risperidone-treated patients improved on domains of visuospatial memory. Haloperidol-treated patients improved only on domains of learning/memory. However, patients able to remain in treatment for the entire 52 weeks benefited more from olanzapine or risperidone treatment than haloperidol treatment.