In general when you have already evaluated your model on unseen data (test set) and its RMSE is different than predictions RMSE, is it ok ? How much difference is fine and how to know that ?
Its fine to have some difference in prediction on training RMSE, Test RMSE and Out of Time sample (Unseen) RMSE. No thumb rule exist right now to say if 5%/10% or any difference is fine. It depends on the problem statement but anything around 5% or above should be investigated properly.
The concept is defined as Overfitting when you do good on training data but model do worse on training data.