ML model artefacts are the outputs and supporting resources generated during model development, often represented in standard formats that support cross-platform portability, allow predictive models to be represented in a vendor-neutral markup, or capture both model and runtime logic. For example, a trained neural network—whether a smaller model or a large language model—can be packaged with its architecture, weights, and configuration files (including hyperparameters, training settings, and data transformations) to facilitate portability and reproducibility across diverse tools and environments.