Hi All, I have seen many posts here mentioning the fact the GPT-4o-mini uses many times more tokens than GPT-4o, such that the cost per image is same. I am wondering how I can count the the number of image prompt tokens ? Currently I am checking the output of GPT-4o-mini for one image, which is as below:
CompletionUsage(completion_tokens=660, prompt_tokens=1098, total_tokens=1758)
While invoking GPT-4o also I am getting the same number of prompt_tokens.
Can someone clarify what I am doing wrong here?
Hi,
This is a known issue and OpenAI are looking into it.