One 4:5 poster, four Sume image models: what a round costs
Flux 2 Pro, Seedream 5 Lite, Qwen Image and Recraft V4 all list 4:5. One image from each costs $0.15625 on Sume. The arithmetic and the catch for each row.

One 4:5 image from each of Flux 2 Pro, Seedream 5 Lite, Qwen Image and Recraft V4 costs $0.0375 + $0.04375 + $0.025 + $0.05 = $0.15625 on Sume. A four-model comparison round is therefore about sixteen cents. All four list 4:5 in their catalog aspect ratios, so the same prompt and ratio go to each without conversion.
The arithmetic
Each price is the provider list rate times 1.25, per the Image API docs, and you pay cost_usd x n. The list rates are in the repo's pricing table.
Note the difference between the headline and the sum: add the four lines exactly rather than rounding each to cents.
| Row | List | Sume price | Edits | Output formats |
|---|---|---|---|---|
| Qwen Image | $0.02 | $0.025 | yes, up to 10 references | png, jpeg, webp |
| Flux 2 Pro | $0.03 | $0.0375 | yes, up to 10 references | png, jpeg, webp |
| Seedream 5 Lite | $0.035 | $0.04375 | yes, up to 10 references | png, jpeg, webp |
| Recraft V4 | $0.04 | $0.05 | no | webp only |
| Round total | $0.125 | $0.15625 |
The catches
Recraft V4 is text-to-image only and returns WebP, so a PNG pipeline needs a conversion step. Qwen Image is the cheapest, which makes it a fair baseline. The cost of four prompts across four rows is 4 x $0.15625 = $0.625.
Do not read price as quality. Sume publishes no ranking of these rows, and we make no claim about which will suit your prompt. The round exists so you can judge by eye.
Run the round
The loop below sends one prompt to all four rows with the same 4:5 ratio and prints each URL. It stops at the first error.
import os, requests
H = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}
MODELS = ["qwen/qwen-image", "black-forest-labs/flux.2-pro",
"bytedance-seed/seedream-5-lite", "recraft/recraft-v4"]
for m in MODELS:
r = requests.post("https://api.sume.com/v1/images", headers=H, timeout=120,
json={"model": m, "prompt": "poster for a night market",
"aspect_ratio": "4:5"})
r.raise_for_status()
print(m, r.status_code, r.json().get("data", [{}])[0].get("url"))Scaling the round
The same round at other volumes is simple multiplication, because each row is a flat price per image. Ten prompts through four models cost 10 x $0.15625 = $1.5625. Adding a fifth row such as Flux 2 Flex at $0.0625 brings one round to $0.21875.
If you only need to cut the field, run the cheap rows first. Qwen Image and Flux 2 Pro together cost $0.0625 per prompt, and you can add Seedream 5 Lite and Recraft V4 only for the prompts where neither convinces you.
- Qwen Image: $0.025
- Flux 2 Pro: $0.0375
- Seedream 5 Lite: $0.04375
- Recraft V4: $0.05
Before you start
Confirm the prices on the endpoint record before you spend anything, even a few cents. Read GET /v1/images/models/{model_id}/endpoints for each row, because list rates move and the catalog is the source of truth for what your wallet is charged. Keep the four generated files side by side, name them after the model id, and note the prompt, so the comparison can be repeated later when a row changes.
Sources
Related posts
More in Comparisons
- SCORM export for AI avatar video: HeyGen, Colossyan, and Sume
HeyGen Business and Colossyan Professional list SCORM export. Plan prices, export limits and quizzes, and what Sume returns instead: a plain MP4 with no SCORM.
- Seedance 2.5 leads 2.0 by 27 Elo on text-to-video, not image-to-video
Seedance 2.5 leads 2.0 by 27 Elo on AA text-to-video and 75 on editing, but 2.0 is fifth on image-to-video and 2.5 is not in the top 12. Sume prices both.
- Seedance 2.5 or 2.0 for a 15 s clip: $8.67 vs $5.67 at 720p
For the same 15 seconds, Seedance 2.5 costs 53% more than Seedance 2.0 at 720p ($8.67 vs $5.67). Where 2.5 earns it and where 2.0 is enough.
- Seedance 2.5 or Wan 3.0 for text-to-video: 13 Elo, 2.8x the price
Wan 3.0 beats Seedance 2.5 by 13 Elo on text-to-video and 36 on editing, at $12.00 against $34.12 a minute. Pick by length, price and inputs.
Written by Sume