Is there any cost difference in using one fine tuned model versus many? In other words, is the cost of sending 100 chat completions to one fine tuned model the same as sending 20 chat completions to each of five fine tuned models? Thank you!
And is there any effective limit to the number of distinct fine tuned models per API key?
Hi!
The costs are determined by the input and output tokens for each fine-tuned base model, so in short, the monetary cost will remain the same.
To reduce costs, you can use less expensive models for your fine-tunes. Additionally, consider exploring model distillation, especially when the questions are narrowly scoped and the answers are concise.
https://platform.openai.com/docs/guides/distillation
As for the maximum number of fine-tuned models, I’m not sure if there’s a limit. However, we haven’t received any reports from other community members about reaching a maximum.
Hope this helps!
Super answer. Thank you very much!