This study analyzed the efficiency of protein refolding with crossflow ultrafiltration for two distinct types of model proteins: denatured bovine alpha-lactalbumin and a fusion protein that comprised green fluorescent protein coupled with an engineered N-pro) autoprotease tag. A mechanistic model of the process dynamics that accounted for the refolding kinetics was developed and verified by comparing with experimental data. The model was used to quantify refolding performance under various operating conditions, including different denaturant removal rates, refolding durations, and protein concentrations. The performance of ultrafiltration with the fed-batch and stepwise operating modes for denaturant removal was analyzed and compared to the performance of the batch dilution method. Performance was evaluated in terms of productivity, yield, and buffer consumption. The superiority of one refolding method over another depended on the protein system. When a slow reduction in denaturant concentration suppressed protein aggregation, the best performance was achieved with the ultrafiltration system. For example, ex-lactalbumin refolding with ultrafiltration achieved several-fold higher productivity and lower buffer consumption compared to refolding with the batch dilution method. A further reduction in buffer consumption was achieved with permeate recycling. Conversely, when rapid dilution of the denaturant was most efficient, a combination of batch dilution and ultrafiltration was recommended. The latter reduced the buffer consumption with permeate recycling; e.g., over 80% of the refolding buffer could be recycled during fusion protein refolding. (C) 2015 Elsevier Ltd. All rights reserved.