6 ms·

Plus, if the LLM goes down (or gets "upgraded" to a model that does the translation differently/wrong), you still have the tool available locally.