C/C++ use python trained model

First, i trained a model in python, and get prediction, use function “save_model” to save the model.
Then I use C/C++ loaded this model, and use the same data to predict, but get different prediction value.
But little data get same prediction value.
python xgboost version is 1.5.2, and C/C++ XGBoostVersion retrun 1.5.1, is the version problem? Or C/C++ should do something after XGBoosterLoadedModel()?