Loading…
33
TIL my fine-tuned model beat the big API for our niche docs
For 3 months I kept calling the GPT and Claude APIs for our legal Q&A bot and got so many generic answers. Last week I finally compared them side by side against a fine-tuned Llama 3 that I trained on 400 of our client transcripts. The fine-tuned model got 87% accurate citations while the APIs were around 60%. The setup took two nights but the difference is night and day for our users. Has anyone else made the jump from generic APIs to fine-tuned models for domain specific stuff? What made it worth it for you?
0 comments
Log in to join the discussion
Log In0 Comments
No comments yet
Be the first to share your thoughts on this discussion.