GPT Image 2.5 on Sume: low 1K $0.02475 vs high 4K $0.2225
GPT Image 2.5 spans 9x in price on Sume: $0.02475 for low at 1K, $0.055625 medium at 2K, $0.2225 high at 4K. Omit quality and the default is high.

GPT Image 2.5 on Sume costs $0.02475 for a low-quality 1K image, $0.055625 for medium at 2K and $0.2225 for high at 4K, a spread of 8.99 times. The Image API default quality is high, so a request that leaves the field out lands at the expensive end.
Ladder and 100-image cost
The three price points come from the Sume catalog, where the catalog default estimate for the model is $0.2225.
| Quality and size | Per image | 100 images | Against low 1K |
|---|---|---|---|
| low, 1K | $0.02475 | $2.475 | 1.00x |
| medium, 2K | $0.055625 | $5.5625 | 2.25x |
| high, 4K | $0.2225 | $22.25 | 8.99x |
The default is the trap
Sume's Image API page says that for openai/gpt-image-2.5 the quality values are auto, low, medium, high, xhigh and max, that omitting it means high, and that auto reserves max. The same page lists fal token rates behind the model: $30 per million output image tokens, $8 per million input image tokens and $5 per million input text tokens, with xhigh at 1024x1024 estimated at $0.09366 and max at $0.21072 before input tokens and Sume pricing.
So the practical rule is to choose quality on every call. A 100-image run left on defaults can cost $22.25 instead of $2.475, a difference of $19.775.
How to use the ladder
Draft at low, finish at medium or high. If a prompt works at low quality, you have bought 90 drafts for the price of 10 high-quality images: 90 x 0.02475 = $2.2275 against 10 x 0.2225 = $2.225. Pay for high only on images that ship.
The model also supports up to 16 image references, an optional mask and a transparent background option, according to the Image API page. Prices can change; the catalog is the current number.
- Always set
quality; the default ishigh. - Draft low, ship medium or high.
autoreservesmax, the top of the scale.
Reproduce the numbers
Every Sume price on this page comes from the public catalog at https://api.sume.com/v1/catalog, which needs no API key. Fetch it, find the row for the job or model named in the table, and multiply by your own seconds, characters or image count. Sume bills the list price shown in that row, so the catalog figure is the one that applies to your balance.
All figures on this page are Sume's own, as of 2026-10-09. Prices and limits can change, so re-read the catalog and the linked docs before a large run, and keep the arithmetic in a spreadsheet so you can update it in a minute.
Sources
Related posts
More in Comparisons
- Griffin-Lite's 0.43 s latency: what a recorded avatar clip gives up
Tavus reports Griffin-Lite video latency of 0.43 s on average (0.27-0.59 s) in a research preview. A recorded clip on Sume is a job, so choose by use case.
- HappyHorse 1.0 at $0.15 a second: four Sume rows cost less at 720p
Runway Dev lists HappyHorse 1.0 at 15 credits per second at 720p and 30 at 1080p. Sume does not list it; four Sume rows are cheaper at 720p, three at 1080p.
- HappyHorse 1.1 lists 9 aspect ratios; Omni on Sume lists 2
HappyHorse 1.1 reference-to-video offers nine ratios, from 9:21 to 21:9. Gemini Omni Flash offers 16:9 and 9:16. How to check a Sume model's ratios first.
- HeyGen Business 5 custom avatars vs five Sume avatars at $4.75
HeyGen Business lists $149 a month plus $20 per seat with 5 custom avatars; creating five avatars on Sume is 5 x $0.95 = $4.75, with video billed separately.
Written by Sume