Fine-tuned model vs. prompt-only setup, which actually wins?

I spent last month testing a custom fine-tuned model for our customer support replies versus just jamming every rule into a long prompt. The fine-tuned version cost $120 to train and took 3 days, but it handled edge cases way better. The prompt-only route was free and quick, but it kept tripping over slang and weird phrasing. Has anyone else gone down this road and found a clear winner for smaller teams around Austin?

1 comments

1 Comment

evannelson1mo ago

Honestly the part nobody talks about is how fine-tuning can lock you into bad patterns if your rules change, like we reworked our return policy and the model kept defaulting to the old one until we retrained it, whereas a good prompt just gets edited in five minutes.