Hi Folks,
I am a little confused about retrieval pricing on files uploaded on the message level.
The document mentions that running an assistant with retrieval enabled on a thread with one message and 1 unique file will incur a per-GB-per-day charge on the file.
My question is will this cost depend on how many times I run the thread with assistant or it will keep adding up whether I run the thread or not?
When I bind a file at the message level within a thread, I’ve noticed that it seems the entire content of the file is also counted towards token consumption during interactions. This means that since each interaction includes the history of the conversation, if the uploaded file is large, the token consumption for each interaction will also be significantly higher, leading to substantial costs. Regarding the mentioned cost of $0.20/GB per assistant per day, this might refer to the expense of OpenAI storing and managing the uploaded file. However, it appears that retrieving the file during conversations also consumes tokens for the entire content of the file. This observation is based on my single experiment. Of course, to verify this, you may want to conduct further tests. I look forward to your response.
1 Like