The models used in cross validation (5, as you mentioned) are not directly combined into a single final model.
The goal of cross-validation is to evaluate the model's performance across different data splits, ensuring it generalizes well. After cross-validation, you discard the five individual models.You train a new random forest on the entire dataset using the best hyperparameters found during cross-validation. This final model is used for predictions.
Refer the following example to read the workflow of training a final model after evaluating and finding hyperparameters through cross-validation: https://www.mathworks.com/help/stats/classificationsvm.crossval.html#mw_e9fd437d-4125-4b9e-b87c-f370bc439a3e
