Grok Imagine aspect ratio: 21:9 and 5:2 on xAI vs Sume
xAI added 21:9 and 5:2 to Grok Imagine image on Aug 28. Which ratios Sume's grok-image accepts, what the 400 looks like, and a call that works.

On xAI's own API, Grok Imagine image takes aspect_ratio values from 1:1 to 21:9, including 5:2, and auto when you leave it out. Sume's x-ai/grok-image accepts 13 ratios: 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, 1:2. 21:9 and 5:2 are not in that list, so a request with either returns a 400.
xAI's numbers come from its image generation docs and release notes; Sume's from the Image API docs and the catalog code. All read 2026-09-29.
Which aspect ratios does xAI document for Grok Imagine?
Aspect ratio is the parameter people trip on first, because a model card and a gateway rarely agree on it. xAI's list moves with its model; a gateway's list moves with its catalog. Compare the two before you copy a request from one into the other.
The release notes for August 28 say image generation and editing now accept 21:9 (cinematic widescreen) and 5:2 (wide banners). The docs table also lists phone-screen ratios 19.5:9 and 20:9 in both orientations, and auto, which lets the model pick a ratio for the prompt.
| Ratio | xAI documents | Sume x-ai/grok-image |
|---|---|---|
| 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3 | Yes | Yes |
| 2:1, 1:2 | Yes | Yes |
| 19.5:9, 9:19.5, 20:9, 9:20 | Yes | Yes |
| 21:9 | Yes (added Aug 28) | No |
| 5:2 | Yes (added Aug 28) | No |
| auto | Yes (default) | No |
| 4:5 | Not in the xAI table | No |
What happens if I send 21:9 to Sume?
Sume validates aspect_ratio against the model's advertised list and returns 400 invalid_request with details.supported holding the accepted values. Nothing is generated and nothing is billed.
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 film still of a lighthouse at dusk",
"aspect_ratio": "21:9"
}'Which wide ratio should I use on Sume instead?
A 400 for an unsupported ratio is cheap to handle: catch it, read details.supported, and retry with the nearest value. Do not retry the same body, since it will fail the same way.
The widest ratios Sume lists for Grok Imagine are 2:1, 20:9 and 19.5:9. Pick 20:9 for the closest thing to a widescreen banner, then crop in your editor if you need exactly 21:9. Sume does not promise exact pixel sizes, so treat a ratio as a shape, not a resolution.
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 film still of a lighthouse at dusk",
"aspect_ratio": "20:9"
}'Does the resolution or n setting differ too?
xAI lists 1k and 2k resolutions and n up to 10. On Sume the catalog for x-ai/grok-image advertises no resolution parameter and one image per call, so a resolution or an n above 1 is rejected rather than ignored. One image is estimated at $0.03 on Sume.
Read the live catalog before you build on any of this:
curl "https://api.sume.com/v1/images/models" \
-H "Authorization: Bearer $SUME_API_KEY"Will Sume add 21:9 and 5:2?
Sume's docs make no promise, so this post makes none. If the catalog changes, GET /v1/images/models shows the new aspect_ratio values for the model.
Sources
Related posts
More in Models
- Grok Imagine Video 1.5 reference images API: xAI vs Sume
xAI's API takes up to seven reference images, text-to-video and 1080p for Grok Imagine Video 1.5. Sume's API reference lists it as first-frame image-to-video.
- Hailuo 3.0 release date: MiniMax H3 launched 2026-07-31
Hailuo 3.0 is the name people use for MiniMax H3, which MiniMax released on 2026-07-31. What the vendor pages call it, and the ids that reach it on Sume.
- How much VRAM for AI video generation? Numbers from model cards
From 14 GB to 80 GB: the VRAM figures Tencent, Wan-AI and Genmo publish for their open video models, and what each number was measured with.
- HunyuanImage 3.0 API: open weights, no listed Sume model
HunyuanImage 3.0 is an 80B-parameter open-weights image model from Tencent. Sume's image docs do not list it; here is what the card says and the hosted route.
Written by Sume