Standard approaches to model calibration typically rely on minimizing the error between model predictions and experimental observations. While straightforward to implement, such methods often neglect uncertainties in the data and may lead to parameter estimates that are overly dependent on the available samples. This limitation becomes particularly critical when data is scarce and model evaluations are computationally expensive or experimentally demanding. Bayesian calibration offers a robust alternative by framing parameter estimation as a probabilistic inference problem. Within this framework, uncertainties arising from experimental noise and model inadequacy are explicitly accounted for, yielding posterior probability distributions over model parameters rather than single-point estimates. Bayesian approaches also naturally incorporate prior knowledge—derived from previous studies, expert judgment, or literature—thereby improving robustness and interpretability under limited data conditions. The resulting uncertainty information can be exploited to assess calibration confidence, guide experimental design, and optimize resource allocation. To enable Bayesian calibration for complex, high-fidelity models, we employ a surrogate-assisted calibration framework based on advanced probabilistic emulators. In particular, Gaussian process–based surrogates are used to efficiently approximate computationally intensive models while preserving uncertainty propagation. The framework extends to multi-output calibration, allowing multiple, potentially correlated experimental observables to be assimilated simultaneously within a coherent Bayesian formulation. This capability is essential for realistic physical models, where different response quantities jointly inform the underlying parameters. The proposed framework is implemented in the open-source Python package ACBICI [1], inspired by the seminal work of Kennedy and O’Hagan [2]. In this talk, we present the underlying methodology and recent enhancements, and demonstrate its practical relevance through the calibration of representative engineering models. We illustrate how advanced surrogates and multi-output Bayesian calibration provide rich uncertainty information and actionable insights for modelers and engineers. [1] M. Kennedy and A. O'Hagan, Journal of the Royal Statistical Society: Series B (Statistical Methodology), 63:425-464, 2001. [2] C. Schenk, I. Romero, GitLab, https://gitlab.com/schenkch/ACBICI, 2026.