Another nice package is Darts. It supports multiple series global models. You mentioned not to use quantiles. Do you mean you do not want quantile statistics for probabilistic forecasting? Monte Carlo, Quantile Regression, Conformal Prediction, and Conformalized Quantile Regression are frequently used techniques for probabilistic forecasting. You can test each of them. You can check out: https://dataman-ai.medium.com/conformalized-quantile-regression-for-time-series-probabilistic-forecasting-85a2a1047119