Protein engineering has been used to enhance the activities, selectivities, and stabilities of enzymes. Frequently tradeoffs are observed, where improvements in some features can come at the expense of others. Nature uses modular assembly of active sites for complex, multi-step reactions, and natural "swing arm" mechanisms have evolved to transfer intermediates between active sites. Biomimetic polyethylene glycol (PEG) swing arms modified with NAD(H) have been explored to introduce synthetic swing arms into fused oxidoreductases. Here we report that increasing NAD(H)-PEG swing arms can improve the activity of synthetic formate:malate oxidoreductases as well as the thermal and operational stabilities of the biocatalysts. The modular assembly approach enables the KM values of new enzymes to be predictable, based on the parental enzymes. We describe four unique synthetic transhydrogenases that have no native homologs, and this platform could be easily extended for the predictive design of additional synthetic cofactor-independent transhydrogenases.