This is the large variant of FinBERT (TurkuNLP/bert-base-finnish-cased-v1). The training data is exactly the same.