Pricing of usage of fine-tuned models

API
1

Hi,
If I am not misinterpreting the pricing table of Fine-tuning models in the pricing page, the usage of a fine-tuned model is 6 times expensive than the usage of original GPT-3.5 base model. Thats is due to the existance of “input usage” and “output usace” coulumns in the table.
Is it so, or not?
If not, what are the meanings of “input usage” and “output usage” in the Fine-tuning models pricing tabe.

Thnks.

2

Hi and welcome!

Your interpretation of the information is correct.

3

The only thing not answered clearly is if the older higher pricing before devday reductions still applies to fine-tune on older model.

If there is no difference in price, then it only makes sense to train the gpt-3.5-turbo-0613 model, as it has much higher quality than the price-reduced recent turbo’s-turbo models.