Yes. The chat functionality is really just a clever trick. All LLMs are still just predicting the next token in a sequence (i.e. finishing the sentence).