Don't believe me? OpenAI themselves use a special-purpose model and not an LLM for moderation. https://arxiv.org/abs/2208.03274

If they don't think this is a situation for an LLM, you can probably safely assume it's not.