Bioprocess development requires the availability of accurate and reliable process models that relate the key operational parameters to the yields. In this paper, the Response Surface Methodology (RSM) and Artificial Neural Network models (ANN) were used to assess the impact of fermentation volume size on process model accuracy. This was performed at two different process scales (80 and 800 mL). Input variables consisted of inoculum size, molasses concentration and hydraulic retention time. ANN-based models gave R-2 values of 0.99 and 0.95 whereas RSM-based models gave R-2 values of 0.97 and 0.89 for 80 and 800 mL, respectively. The experimental validation of the optimized models gave yields of 0.89 and 0.71 mol H-2/mol sucrose consumed (ANN models) compared to 0.99 and 0.70 mol H-2/mol sucrose consumed (RSM models) for 80 and 800 mL, respectively. These models showed relatively negligible deviations from their predicted values across both process volumes. Semi-pilot scale (8 L) process assessments under optimized conditions showed negligible yield discrepancies from the predicted values of the models at flask scale. These findings suggested that miniaturization of experiments does not significantly impact on the model accuracy, thus reducing costs during the process developmental stage. (C) 2017 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.