I hope it's just that and the costs quickly go down. Seeing the cost of gpt-3.5-turbo, and since many rumors have stated that GPT-4 isn't much larger than GPT-3.5, I wasn't expecting such a high price.
It's not the size of the GPT, but the magic in it that makes the difference....
Could be that while the number of parameters are not significantly more, they could be doing more computation per token.