I was hoping to fine tune gpt-4o-mini on my LLM game i’m making, so it can understand the proper ways to process game logic (mostly just what tools to call and in what scenarios with what inputs).
The thing is, I don’t have the money to be hosting an LLM game for others to play on my dime. So I was thinking of creating a simple “login” interface where users simply supply their own api key and the game is playable using their OpenAI credits. Players could use their $5 trial credits that you get when creating an account to play the game.
It would be nice to be able to user other’s api keys, but give them access to my fine tuned version of 4o-mini, for the best player experience (also cheaper on tokens which would let me experiment with other interesting ways to make the game more immersive).
If not I think I’ll have to stick with gpt-4o and hope the base model can comprehend and not be too expensive (it seems to not be too expensive, but im not sure if it can comprehend how to effectively process game logic)