An assessment framework is developed that allows analysts to conduct stock assessments for fish and invertebrate stocks based on age-, size- and age-size-structured population dynamics models. The size-structured model is nested within the age-size-structured model. The framework can use catch, discard, index of abundance, size- and age-composition, conditional age-at-length, mean length-at-age, and tagging data to estimate model parameters. It is used to explore the sensitivity of key model outputs for Pacific cod in the Eastern Bering Sea by applying model configurations that use the same data, same likelihood functions, and same data weighting schemes. Base model configurations using the three model types all fit the available data adequately, but the age-structured model fits the data better than the size-structured model. Variation in estimates of spawning biomass and the overfishing level was higher among model-types than within model-types. This result highlights the need for assessment analysts to focus more on applying and presenting results for multiple models.