I was working with an assistant and at first it was very cheap an then it got very expensive. It seems like the uploads in a chat are not only not remembered between threads but also much more expensive? Is the cost of uploading when creating or updating an assistant cheaper or free compared to uploading to an assistant during a conversation?
Hi and welcome to the Developer Forum!
Any files that get uploaded have to be processed, i.e. vectorised and stored, that will cost tokens. Once files are uploaded they can be reused, and that will only use the token count required to perform the retrieval which will typically be smaller.