Model returning different values in python and scala

I have a model trained using “objective:ndcg” and it is returning different values in python and scala.
In Python I’m using
model = xgb.Booster()
model.load_model("…")

And in Scala I’m using
val model = XGBoost.loadModel("…")

When I call model.predict() on an identical array, it gives me 2 different numbers.
Is this possible? Am I doing something wrong?

UPDATE: I’m not doing anything wrong, this seems like a pretty big issue. Unfortunately I can’t share the model since it is proprietary but I can help reproduce this issue if someone is looking into it

Pay close attention to how missing values are represented in your input array. Also, the array should be of 32-bit float type.