How are image costs calculated

[profoundvision]

Hi,

Based on what I can find in the docs and code, and my experience with GPT-image and gemini-image models, the cost of images is calculated per image. But for the multimodal generation model the actual price is determined based on image and text tokens.
Is the price calculated on these tokens or per image?

[profoundvision]

Any updates?

[profoundvision]

While previously I saw a fixed price per image, it now seems to be dependent on tokens, but the calculation is off. So filing a bug report.