Recraft V4 vs FLUX.2 Pro on Sume: $0.05 text-only vs $0.0375 with refs
On Sume, Recraft V4 bills $0.05 per image, is text-only and returns webp only. FLUX.2 Pro bills $0.0375 and takes 10 references. Both list 13 ratios.

On Sume, Recraft V4 costs $0.05 per image, takes no reference images and returns webp only, while FLUX.2 Pro costs $0.0375, takes up to 10 references and lists png, jpeg and webp. Both list 13 aspect ratios. So FLUX.2 Pro is the cheaper and more flexible row on contract terms, and Recraft V4 is for cases where you have seen its output and want it.
Sume figures are from the catalog code and the Image API docs, read on 2026-10-10. The fal FLUX.2 Pro page, read that day, lists $0.03 for the first megapixel and $0.015 per extra, which Sume flattens to a single billed price.
Contract comparison
Neither row lists resolution, quality, mask_url or background. The differences are in price, references and output format.
| Setting | Recraft V4 | FLUX.2 Pro |
|---|---|---|
| Billed per image | $0.05 | $0.0375 |
| References | none | up to 10 |
| Aspect ratios | 13 | 13 |
| Output formats | webp only | png, jpeg, webp |
What the webp-only rule means
If a design tool, ad platform or print vendor needs png or jpeg, a webp-only row needs a conversion step, which is easy in code but one more thing to run and check. Do not send output_format: "png" to the row, since the descriptor lists only webp and anything else is a 400.
FLUX.2 Pro accepts all three formats, so the same request body can ask for a png when you need alpha-free lossless output. This does not create transparency: the docs name only the GPT Image 2.5 background field as a route to a transparent generated image.
Cost at volume and a decision rule
A thousand images cost $50 on Recraft V4 and $37.50 on FLUX.2 Pro, a $12.50 gap. That is small. The reference limit is the larger difference: Recraft V4 cannot start from your product photo or style board.
Generate five prompts on each row, hide the labels and pick. If you pick Recraft V4 for more than half, use it for text-to-image brand art and keep FLUX.2 Pro for anything that needs a reference.
- Need references: FLUX.2 Pro.
- Need png or jpeg without conversion: FLUX.2 Pro.
- Prefer Recraft V4's look after a blind test: use it, text-only.
- Pin the row id for repeatability.
Using each in a brand workflow
A brand workflow usually has two halves: exploration from words, and production from existing assets. Exploration can use either row, and the cost per image is close. Production needs references, so it runs on FLUX.2 Pro or another row with a reference limit above zero.
Keep one prompt library and tag each prompt with the row it was tested on, so a change in row does not quietly change the look of a campaign. Save the row id with every output and the date of the test.
Checks before a large run
Read the endpoints route for both rows and confirm the prices in the table above, then generate a small batch of each at the ratios you need. Confirm the file type that comes back, open a few outputs in the tools your team uses and make sure the webp files open there. Those fifteen minutes cost less than a batch regenerated in a different format.
If you do pick Recraft V4, state the webp requirement in the job brief so downstream steps expect it, and keep a conversion step in the pipeline for anything that has to leave as png or jpeg. A small, explicit step is better than discovering the format at the end of a batch.
Finally, re-read the catalog before you commit. The row prices and the reference limits are the facts that drive the decision, and both are published by the list and endpoints routes at the moment you call them, which is better than a number in a blog post. This comparison was read on 2026-10-10 and should be refreshed when either row changes.
Sources
Related posts
More in Comparisons
- Remotion Automator $0.01 a render vs Sume Timeline $0.10 a minute
Remotion lists $0.01 a render with a $100 monthly minimum; Sume Timeline bills $0.10 per output minute. Break-even by volume and what each leaves out.
- Remotion Lambda or a Sume Timeline job: who runs the servers?
Remotion Lambda renders in your AWS account; a Sume Timeline render is one billed job on Sume's workers. What you operate, and the limits, read 2026-10-10.
- Remotion or Sume Timeline: React code or a JSON document?
Remotion defines video as React code; Sume Timeline 1.0 assembles Sume-hosted clips from a JSON document. Where each fits, with limits read on 2026-10-10.
- Runway's API lists 19 video ids: which ones have a Sume id?
Of 19 video ids on Runway's pricing page, 9 map to a Sume id by name, 2 are unconfirmed, and 8 have no Sume equivalent. Full mapping table with the Sume ids.
Written by Sume