Image edits with gpt-image-1.5 in the playground are billed but do not appear in usage dashboard

[cmoi]

Hi,

I have activated a credited an API account to evaluate the use of gpt-image-1.5, I have tested some edit prompts in the playground and wanted to check the cost of each one, but in the usage dashboard I can only see the total cost of the sum of all calls.
I am not able to see how many image calls, which day, using which key, nothing.
Is it something that happens to new accounts ? Is it only happening because gpt-image-1.5 is new and it does not yet appear in the dashboard ? Is there a delay (more than two days?) before the dashboard gets updated ?

I ended up stopping my tests as I would need to do one test, get last value, go to dashboard, get new value then see the difference to get an idea of the cost of each image edit prompt.

[_j]

You can see the edit request, but the location is bizarro-land: in chat completions:

Either it is un-thought or it is purposeful: because the model is basically text+multimodal now, with prompt text and images as input, and reasoning text and image as output.

Also noteworthy is the input image should have vision capped at 512x512 = 32x32 patches = 1024 tokens if it is patches (while gpt-image-1 is “tiles” and 1 tile is 129+65=194 tokens) , but this has blown right past. It’s not 20% cheaper, it’s 20x+ the token cost for input.