GPT Image 2.5 high: 1920x1080 costs less than 1024x1024

On fal, GPT Image 2.5 Flare at high quality is $0.0396 for 1920x1080 and $0.05268 for 1024x1024, so the wider frame costs about 25 percent less despite having twice the pixels. The price follows output tokens, and the 1080p frame uses fewer than the square frame. Sume bills these rows from the same fal token rates, so check the job cost rather than assuming price rises with size.

The fal numbers

The fal page (fetched 2026-10-07) lists the three prices below at high quality, which is its default. It adds that longer prompts, more complex requests and larger images cost more.

What Sume does with it

The Sume Image API docs say Flare and Sunburst use the same fal token rates: $30 per million output image tokens, $8 per million input image tokens and $5 per million input text tokens. The catalog list price is high quality at 1024x1024 only, and an estimate for other sizes is made per request. The billed amount comes back in usage.cost on the response, and endpoint pricing already includes the Sume margin.

So a 16:9 request does not have to cost more than the square default. It also means the catalog list price is not a ceiling for every size.

How to use this

Do not infer cost from pixel count. Run one request per size you ship and read usage.cost.

• Use aspect_ratio 16:9 for slides and thumbnails; it stays under the square price on fal.
• Pin quality explicitly. Omitting it means high, and auto reserves the maximum price on Sume.
• Remember 3840x2160 is a custom size: both edges must be multiples of 16, and the 3:1 aspect limit applies.