A bench scale cell culture model representative of manufacturing scale (2,000 L) was developed based on oxygen mass transfer principles, for a CHO-based process producing a recombinant human protein. Cell culture performance differences across scales are characterized most often by sub-optimal performance in manufacturing scale bioreactors. By contrast in this study, reduced growth rates were observed at bench scale during the initial model development. Bioreactor models based on power per unit volume (P/V), volumetric mass transfer coefficient (k(L)a), and oxygen transfer rate (OTR) were evaluated to address this scale performance difference. Lower viable cell densities observed for the P/V model were attributed to higher sparge rates and reduced oxygen mass transfer efficiency (k(L)a) of the small scale hole spargers. Increasing the sparger k(L)a by decreasing the pore size resulted in a further decrease in growth at bench scale. Due to sensitivity of the cell line to gas sparge rate and bubble size that was revealed by the P/V and k(L)a models, an OTR model based on oxygen enrichment and increased P/V was selected that generated endpoint sparge rates representative of 2,000 L scale. This final bench scale model generated similar growth rates as manufacturing. In order to take into account other routinely monitored process parameters besides growth, a multivariate statistical approach was applied to demonstrate validity of the small scale model. After the model was selected based on univariate and multivariate analysis, product quality was generated and verified to fall within the 95% confidence limit of the multivariate model. (C) 2014 Wiley Periodicals, Inc.