NAD(+) and NADP(+), chemically similar and with almost identical standard oxidation-reduction potentials, nevertheless have distinct roles, NAD(+) serving catabolism and ATP generation whereas NADPH is the biosynthetic reductant. Separating these roles requires strict specificity for one or the other coenzyme for most dehydrogenases. In many organisms this holds also for glutamate dehydrogenases (GDH), NAD(+)-dependent for glutamate oxidation, NADP(+)-dependent for fixing ammonia. In higher animals, however, GDH has dual specificity. It has been suggested that GDH in mitochondria reacts only with NADP(H), the NAD(+) reaction being an in vitro artefact. However, contrary evidence suggests mitochondrial GDH not only reacts with NAD(+) but maintains equilibrium using the same pool as accessed by β-hydroxybutyrate dehydrogenase. Another complication is the presence of an energy-linked dehydrogenase driving NADP(+) reduction by NADH, maintaining the coenzyme pools at different oxidation-reduction potentials. Its coexistence with GDH makes possible a futile cycle, control of which is not yet properly explained. Structural studies show NAD(+)-dependent, NADP(+)-dependent and dual-specificity GDHs are closely related and a few site-directed mutations can reverse specificity. Specificity for NAD(+) or for NADP(+) has probably emerged repeatedly during evolution, using different structural solutions on different occasions. In various GDHs the P7 position in the coenzyme-binding domain plays a key role. However, whereas in other dehydrogenases an acidic P7 residue usually hydrogen bonds to the 2'- and 3'-hydroxyls, dictating NAD(+) specificity, among GDHs, depending on detailed conformation of surrounding residues, an acidic P7 may permit binding of NAD(+) only, NADP(+) only, or in higher animals both.