Large-scale glacier evolution models are widely used to generate projections of glacier mass change at regional- to global-scales. In model intercomparison projects, these projections come from multiple different models, allowing for the uncertainties associated with different model structures to be assessed. However, these intercomparisons tend to ignore the uncertainties associated with poorly constrained parameters. Therefore, these projections may miss important contributions to uncertainty, but we lack estimates of the size of these uncertainties. To bridge this gap, we quantify parametric uncertainty in projections of glacier volume change in Iceland under experiments from the glacier model intercomparison exercise, GlacierMIP3. To do so, we perform experiments with a large-scale glacier evolution model, ‘GO-VA’, using an ensemble of calibrated parameter sets, rather than with a single set of model parameters as was the case in GlacierMIP3. Our results show that parametric uncertainty can be a major, and in some cases dominant, source of uncertainty in projections of glacier volume change. We find that failing to account for parametric uncertainty reduces overall projection uncertainty by 7–91 % across scenarios of global mean temperature change, with the largest reductions occurring for scenarios where climate forcing uncertainty is highly constrained. Comparison with the GlacierMIP3 ensemble suggests that parametric uncertainty is comparable to structural model uncertainty and, depending on the strength of the forcing, can even be larger. Taken together, these findings highlight the importance of accounting for parametric uncertainty, alongside structural model uncertainty, in model intercomparison projects to more comprehensively characterise projection uncertainty.