Sceptical that an LLM would produce good literary translations - that requires a nuanced understanding of the author's rhythm and style. Maybe if the model was trained only with writing from that one particular author.