4:5 image models on Sume: the portrait ratio and prices
Fifteen Sume image models list 4:5, from Qwen Image at $0.025. See the list, the ones that skip it (Grok Imagine, Imagen 4), and a request.
4:5 is a common feed portrait ratio, and fifteen of the 19 Sume image models list it (read 2026-10-03): GPT Image 2, 2.5 and 2.5 Sunburst, Nano Banana 2 and Pro, the three Seedream rows, Qwen Image and Max, Flux 2 Pro and Flex, Recraft V4, and both Ideogram rows. The four that do not are Grok Imagine, Imagen 4 Fast and Ultra, and Higgsfield Soul. For those, ask for 3:4 and crop.
The 4:5 list
The list comes from the catalog descriptors and the per-image endpoint prices (list x 1.25), read 2026-10-03.
| Model | Price per image | Edits with image_urls | n per call |
|---|---|---|---|
| Qwen Image | $0.025 | yes | 4 |
| Seedream 4 | $0.0325 | yes | 4 |
| Flux 2 Pro | $0.0375 | yes | 4 |
| Seedream 5 Lite | $0.04375 | yes | 4 |
| Seedream 4.5 | $0.05 | yes | 4 |
| Recraft V4 | $0.05 | no | 4 |
| GPT Image 2.5, high, 1024x1280 (estimate) | $0.0566 | yes | 4 |
| Flux 2 Flex | $0.0625 | yes | 4 |
| Ideogram V3 | $0.075 | yes | 4 |
| Nano Banana 2 | $0.10 | yes | 4 |
Which models skip it
Grok Imagine lists 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 but not 4:5, and Imagen 4 stops at 1:1, 16:9, 9:16, 4:3 and 3:4. A model only accepts the ratios its catalog descriptors list, so ask for 3:4 and crop one-sixteenth from the height.
The GPT Image 2.5 case
GPT Image 2.5 is the model with a documented 4:5 pixel mapping in the repo's estimate table: 1024 by 1280. Its price depends on quality, so a 1024 by 1024 render is estimated between $0.0074 at low and $0.2635 at max, and a 1024 by 1280 render at high is estimated at $0.0566. Check the live price before a batch.
A request
Four 4:5 takes on Seedream 4 cost $0.13.
curl -s https://api.sume.com/v1/images \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "bytedance-seed/seedream-4", "aspect_ratio": "4:5", "n": 4, "prompt": "Editorial portrait of a ceramicist at her wheel, window light, warm tones, shallow depth of field"}' What Sume does not do
It does not resize to a platform pixel count. Download the file, check its size, and resize with Pillow if a slot demands exact pixels.
Related posts
More in Models
- Cheap AI video drafts: which models start at 480p or lower on Sume
Seedance, Wan 3.0, MiniMax H3 and Grok Imagine start at 480p, Gemini Omni Flash at 360p, Kling 3 at 720p. Where a draft is possible before a 1080p final.
- 6-second AI video API: every Sume model that accepts 6 seconds
All ten prompt-driven Sume video rows accept a 6-second request. A cost table for Wan 3.0, MiniMax H3, H3 Max, Gemini Omni Flash 1.1 and Kling 3.0.
- 8-second AI video API: cost by model on Sume (Auto's default)
Eight seconds is sume/auto's default and sits inside all ten prompt-driven windows. Billable cost per model at its lowest and highest resolution.
- ACE-Step 1.5 vocal-to-background music: what Sume does instead
ACE-Step 1.5 lists a vocal-to-background-music mode, covers and repainting under MIT. Sume's music jobs are prompt-only, so here is how to get a vocal-free bed.
Written by Sume