Ask HN: Have you fined-tuned any GPT models?

What does you fine-tuned model do and how much did it cost to tune?