Grok Imagine image API: the x-ai/grok-image id on Sume

Sume lists Grok Imagine as x-ai/grok-image on its image API: thirteen aspect ratios (no 4:5), one reference image for edits, and a $0.025 price line per image.

4 min readSume
All posts

To call Grok Imagine for images through Sume, send x-ai/grok-image as the model on POST /v1/images. The catalog names it "Grok Imagine", gives it 13 aspect ratios (none is 4:5), and shows $0.025 per image on the price line. It accepts a text prompt, and edits take one reference image.

This comes from the Image API docs and the catalog code, read 2026-09-29. It says nothing about how the model was built or how it compares with others.

What does the Grok Imagine entry list?

The catalog marks the model as able to do both text-to-image and image edits, with png, jpeg and webp outputs. For edits, the router catalog caps reference images at 1, where most edit-capable entries allow 4.

x-ai/grok-image in the Sume catalog, read 2026-09-29.
ItemValue
Model idx-ai/grok-image
Catalog nameGrok Imagine
ModesText-to-image and edit
Reference images in the router catalog1
Output formatspng, jpeg, webp
Price line$0.025 per image

Which aspect ratios can I ask for?

Thirteen: 2:1, 20:9, 19.5:9, 16:9, 4:3, 3:2, 1:1, 2:3, 3:4, 9:16, 9:19.5, 9:20 and 1:2. There is no 4:5, so a feed portrait crop from this model needs 3:4 or 2:3 and a trim afterwards. A value outside the list is refused as 400 unsupported_parameter.

Does it take a quality or resolution setting?

Sume's docs name resolution only for a few other models, and the SDK's quality hint mentions Ideogram only. Do not send either field to x-ai/grok-image. Read supported_parameters from its endpoint record if you want the definitive list, as the docs advise.

How do I call it?

A plain text request needs the model, a prompt and, if you care, a ratio from the list above.

curl -X POST https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "x-ai/grok-image",
    "prompt": "A rooftop cafe at dusk, wide shot",
    "aspect_ratio": "20:9"
  }'

How many images can one request return?

Sume documents n as 1 to 10 in general, with a lower ceiling for each model, and the older router catalog uses a 1 to 4 range for most image models. Because the ceiling for this entry is not spelled out in the prose docs, ask for one image first and read the endpoint record before you batch. Whole-request behaviour is simple: an image job completes or fails as a unit, so a request never returns a partial set that you would pay for.

When a call runs past 30 seconds, POST /v1/images returns a 202 job envelope instead of blocking, and you poll the job as you would for any other image model.

What does a Grok Imagine image cost?

The price table holds $0.02 list, and Sume's 1.25 margin gives the $0.025 price line. The catalog's billing formula is "list × 1.25 → ceil usd cents", so one image reserves $0.03. Failed or cancelled generations are not billed. The per-model math is laid out in how Sume prices image API calls, and reference-image rules are in image generation with reference images.

Sources

Related posts

More in Models

All Models posts

Written by Sume