Is it possible to reduce the cost and speed of classification endpoint while keeping the performance?

Using classification with a file-dataset, if we increase the number of samples, would it reduce the cost / speed somehow?

To my understanding, with more samples, the search will be faster right? But then I’m not sure how the model pick “selected examples”, because I think more selected examples means higher cost & inference time.

TLDR: How to reduce classification price and speed while keeping / improving performance?

The best way to improve speed and cost is using fine tuning. OpenAI API


I understand, but we had much better results with the classification endpoint than with fine-tuning since we had little amount of data

1 Like