Enzyme-constrained genome-scale models (ecGEMs) have potential to predict phenotypes in a variety of conditions, such as growth rates or carbon sources. This study investigated if ecGEMs can guide metabolic engineering efforts to swap anaerobic redox-neutral ATP-providing pathways in yeast from alcoholic fermentation to equimolar co-production of 2,3-butanediol and glycerol. With proven pathways and low product toxicity, the ecGEM solution space aligned well with observed phenotypes. Since this catabolic pathway provides only one-third of the ATP of alcoholic fermentation (2/3 versus 2 ATP per glucose), the ecGEM predicted a growth decrease from 0.36 h−1 in the reference to 0.175 h−1 in the engineered strain. However, this <3-fold decrease would require the specific glucose consumption rate to increase. Surprisingly, after the pathway swap the engineered strain immediately grew at 0.15 h−1 with a glucose consumption rate of 29 mmol (g CDW)−1 h−1, which was indeed higher than reference (23 mmol (g CDW)−1 h−1) and one of the highest reported for S. cerevisiae. The accompanying 2,3-butanediol- (15.8 mmol (g CDW)−1 h−1) and glycerol (19.6 mmol (g CDW)−1 h−1) production rates were close to predicted values. Proteomics confirmed that this increased consumption rate was facilitated by enzyme reallocation from especially ribosomes (from 25.5 to 18.5 %) towards glycolysis (from 28.7 to 43.5 %). Subsequently, 200 generations of sequential transfer did not improve growth of the engineered strain, showing the use of ecGEMs in predicting opportunity space for laboratory evolution. The observations in this study illustrate both the current potential, as well as future improvements, of ecGEMs as a tool for both metabolic engineering and laboratory evolution.