Is there a good benchmark comparing multilingual and/or translation abilities of most recent LLMs? GPT-4o struggles for some tasks in my language learning app.