Cheapest AI image edit per reference slot on the Sume catalog

If you edit with many reference images (product, logo, model, backdrop), the useful number is reference slots per dollar. On the Sume catalog Qwen Image and Grok Imagine tie at 400 slots per dollar (10 references at $0.025), while ChatGPT Image 2.5 gives the most slots per call (16) at $0.0659.

The figures below come from GET /v1/images/models (the input_references range) and the per-model endpoint price, both read on 2026-10-03. Slots per dollar is only a budgeting ratio: it says nothing about how well a model uses each reference.

Slots per dollar, edit-capable models

Models whose input_references range is 0 to 0 (Soul, Qwen Image Max, Imagen 4 Fast and Ultra, Recraft V4) are text-to-image only and are left out. The price is what one image costs on Sume, with the house margin already applied.

Read the ratio with three caveats

The first caveat is quality: a cheap model with ten slots is not a cheaper ChatGPT Image 2.5 with ten slots. The second is the shape of the call. Grok Imagine accepts ten references but n is capped at 1, so a four-variant plan needs four calls. The third is Ideogram 4.5, which edits the first image and treats up to four more as references, so its five slots are one subject plus four guides.

GPT Image 2.5 prices on Sume are the catalog's high-quality 1024 square rate; larger sizes, xhigh and max cost more, and input image tokens are billed on top.

• Need 11 to 16 references: only the two ChatGPT Image 2.5 ids qualify.
• Need 10 references at the lowest price: Qwen Image or Grok Imagine.
• Need exactly one subject plus a few style guides: Ideogram 4.5 at 5 slots is enough.

A three-reference request

Each reference is an image_url object with a public HTTPS URL; localhost, private-network and non-HTTPS URLs are rejected before submission. The call below blocks up to 30 seconds and returns 200; if the generation outlasts that window you get a 202 job envelope instead, so check the status code rather than the body shape.

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": "Place the bottle from image 1 on the marble from image 2 with the lighting of image 3",
    "aspect_ratio": "auto",
    "input_references": [
      {"type": "image_url", "image_url": {"url": "https://example.com/bottle.png"}},
      {"type": "image_url", "image_url": {"url": "https://example.com/marble.jpg"}},
      {"type": "image_url", "image_url": {"url": "https://example.com/light.jpg"}}
    ]
  }'