GPT Image 2.5 at 16:9 costs 54 to 56% of a square at the same quality

At the same quality, a 1280x720 ChatGPT Image 2.5 frame bills 54 to 56 percent of a 1024x1024 frame on Sume: $0.004 against $0.0074 at low, $0.0356 against $0.0659 at high, $0.142 against $0.2635 at max. The model is billed on output image tokens, and a 16:9 frame at that size uses fewer of them. The Image API docs give the token rates: $30 per million output image tokens, before Sume's margin.

If you generate wide thumbnails or tall covers on this model, the saving is automatic, provided you ask for explicit pixels.

What is the full price ladder?

These are billable prices for one text-to-image call with no references, computed from Sume's catalog estimate and rounded as the catalog rounds. The last column is the 16:9 price as a share of the square.

Why does this only work with explicit pixels?

Sume documents that named presets without a verified GPT-specific pixel mapping reserve the upper bound of output tokens, and that auto quality reserves max. If you send aspect_ratio alone, you may be reserving more than the table shows. Send image_size: {"width":1280,"height":720} and a named quality, and read usage.cost on the response to see what you were billed.

What should I do with this?

For a thumbnail batch, draft at low and finish at high at 1280x720; a 16:9 high frame costs about 2.2 times a medium square. Keep square art for square slots. Prices here are list x 1.25 in dollars per image before whole-cent rounding; the catalog's billable formula reads "list x 1.25, ceil usd cents", so confirm the first charge in usage.cost and budget from the invoice, not from the sum.