Model and simulation-based optimization has emerged as a powerful methodology in science, engineering, and industrial systems. A recurring fundamental challenge is selecting appropriate model fidelity that balances accuracy against computational cost and the resulting constraints on optimization iterations under limited resources. Contrasting and complementary to dynamic multi-fidelity approaches, this work focuses on selecting a single fidelity level per optimization run. This work presents a statistical meta-process modeling framework for analyzing accuracy-cost trade-offs to support offline a priori fidelity or budget selection in model-based optimization. The framework combines three components: a modeling error model that represents accuracy as a function of fidelity, a cost model that describes computational expense per evaluation, and an optimization error model that characterizes optimality-gap convergence across iterations. The corresponding model parameters can be estimated from empirical data obtained from previous experiments on representative problems. An open-source Python implementation is provided for numerical approximations of optimal-fidelity curves and budget-performance plots. Results from case studies demonstrate that the framework can serve as a practical tool for fit-for-purpose model selection in budget-aware optimization settings.