Nano Banana Pro or GPT Image 2.5 xhigh and max at 1K: price on Sume

On Sume, ChatGPT Image 2.5 xhigh is cheaper than Nano Banana Pro at 1K, and max is dearer. At 1024x1024 the three cost $0.117125 for xhigh, $0.1875 for Nano Banana Pro at 1K and $0.2635 for max. Pro sits 60 percent above xhigh and max sits 41 percent above Pro. If your job is a premium hero image, these are the three top-end rows to test against each other.

The comparison is limited to what Sume lists and bills. It says nothing about which output looks better, which you can only settle by rendering your own prompts.

The three rows

Nano Banana Pro has four resolution tiers and no quality field. ChatGPT Image 2.5 has quality and no resolution; its size is set by the aspect ratio and a short-side grid. The table lines up the top end of each.

What moves the number

For Nano Banana Pro, price follows resolution: choose 1K and you pay the 1K price, choose 4K and you pay more. For ChatGPT Image 2.5, price follows quality and the ratio. Reducing the ratio's short side lowers the token count, so 4:5 costs less than 1:1 at the same tier. Setting quality: "auto" reserves the max price on Sume, so set the tier explicitly.

A batch of 20 final images costs $2.34 at xhigh, $3.75 at Pro 1K and $5.27 at max.

Edits and references

All three rows take reference images, so the same prompts work for a retouch or a restyle. ChatGPT Image 2.5 lists up to 16 references and Nano Banana Pro up to 10. If your workflow stacks many references, such as a product, a logo and three style images, the 16-slot ceiling matters more than the price gap.

Both vendors' models are priced in Sume's catalog on the same billed-equals-list-times-1.25 basis, so the table is like for like. There is no surcharge for edits on either row in the figures above, but reference images on ChatGPT Image 2.5 add input tokens, which the calculator prices at $8 per million image input tokens.

Resolution against quality, in plain terms

The two models ask different questions. Nano Banana Pro asks how big you want the image: 0.5K, 1K, 2K or 4K, and you pay by size. ChatGPT Image 2.5 asks how much effort you want spent on it: low, medium, high, xhigh or max, and you pay by effort and shape. A 1K Nano Banana Pro image is not the same thing as a 1024 square at xhigh; the table matches them only by price and rough size.

That matters if you need a large file. Nano Banana Pro lists a 4K tier. ChatGPT Image 2.5 does not list resolution, so ask Sume for the exact size with image_size if you need one, and check the price on the /endpoints record for that size before you run a batch.

When to pay for max

Max is $0.1464 more per image than xhigh. Over 20 images that is $2.93. If you are rendering a poster-size hero or a product image that will be zoomed, a test of xhigh against max on five of your own prompts costs $0.59 and $1.32, and tells you whether the step is visible to you. If it is not, xhigh saves 56 percent.

Request shape

Only the model id and one tier field differ between the calls.

Keep in mind that every figure here is a billed price from Sume's catalog on the date in the table caption. Prices and limits can change, so before a large run, read the endpoint record for the exact model id and compare it with your plan. A one-minute check costs nothing, and it is the only way to be sure the number in your budget is the number on the invoice.

curl -X POST https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-image-2.5",
    "prompt": "Hero photo of a ceramic mug on oak, window light",
    "aspect_ratio": "1:1",
    "quality": "xhigh"
  }'