200 SKU product images on GPT Image 2.5: low $2, medium $4, high $14
Priced per image on Sume: 200 product images on gpt-image-2.5 cost $2.00 at low quality, $4.00 at medium and $14.00 at high. Cheaper and dearer models compared.

Two hundred product images on GPT Image 2.5 through Sume cost about $2.00 at quality: low (1 cent each), $4.00 at medium (2 cents) and $14.00 at high (7 cents, for 1024-class output), before any retries (read 2026-10-08). GPT Image 2.5 shipped as ChatGPT Images 2.5 on 2026-09-08, and Sume lists it as openai/gpt-image-2.5 (Flare) and openai/gpt-image-2.5-sunburst.
How the per-image price is made
Sume bills the catalog list price times 1.25 and rounds up to a whole cent per request. The catalog list for gpt-image-2.5 is the high-quality 1024 output at $0.0527, so high is $0.0527 x 1.25 = $0.0659, which rounds up to 7 cents. Sume's catalog data (read 2026-10-08) puts low at 1 cent and medium at 2 cents at 1024-class sizes. Admission also counts estimated input tokens, so a request with many reference images can reserve more than the display rate.
| Model | Catalog list per image | Billable per image | 200 images |
|---|---|---|---|
| gpt-image-2.5 (low) | not shown (per-quality) | 1 cents | $2.00 |
| gpt-image-2.5 (medium) | not shown (per-quality) | 2 cents | $4.00 |
| gpt-image-2.5 (high) | $0.0527 | 7 cents | $14.00 |
| nano-banana-2.1 | $0.0800 | 10 cents | $20.00 |
| seedream-5-lite | $0.0350 | 5 cents | $10.00 |
| flux-2-pro | $0.0300 | 4 cents | $8.00 |
| grok-image | $0.0200 | 3 cents | $6.00 |
What the model accepts for catalog work
GPT Image 2.5 accepts up to 16 image references, an optional mask_url, and background: auto|transparent|opaque. If you omit quality the default is high, so a catalog job that forgets the field costs 7 cents, not 1. auto quality reserves max, so set low or medium explicitly for bulk work.
Custom pixel sizes need both edges to be multiples of 16, a longest edge of 3,840, a ratio of at most 3:1, and between 655,360 and 8,294,400 pixels.
- Default quality when omitted: high
autoquality reserves max- Up to 16 image references; optional mask
A low-cost loop
Generate one image per SKU at low quality, keep the ones that pass your review, and re-run only the keepers at high. If one in four needs the upgrade, 200 x 1 cent + 50 x 7 cents = $5.50 against $14.00 for all-high. Review the low drafts for label text and edges before you spend the 7 cents.
curl -X POST https://api.sume.com/v1/images \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: sku-0001-draft" \
-d '{
"model": "openai/gpt-image-2.5",
"prompt": "Product on a seamless white background, soft shadow, centered, packshot",
"quality": "low"
}'Sources
Related posts
More in Use cases
- 21 Shorts a week at 59 seconds: render and captions cost $6.30
Three Shorts a day for a week, each held to 59 seconds: one Timeline render ($0.10) and one caption job ($0.20) per Short comes to $6.30.
- 250 product photos to 5-second 1080p clips with Omni: $234.38 on Sume
250 SKUs as 5-second image-to-video clips cost $234.38 at 1080p, $156.25 at 720p and $46.88 at 360p on Sume's Omni row. A draft plan brings it to $103.13.
- 28-minute podcast episode on Sume: narration, bed, render = $4.125
A 28-minute narrated episode costs about $4.125 on Sume: two TTS requests ($1.20), one Music 1.0 bed ($0.125) and a 28-minute Timeline render ($2.80).
- Vidu Q4 takes 3 audio references; which Sume video models take one?
Sume does not list Vidu Q4 Preview. Seedance 2.x, Wan 3.0 and MiniMax H3 accept reference audio; Gemini Omni Flash 1.1 does not. Prices for 15 seconds.
Written by Sume