FLUX 3 Image reference size: 256 px to 16 MP, and Sume URL rules
BFL says each FLUX 3 Image reference is 256x256 px to 16 MP, 1 to 10 images. Sume lists FLUX.2 ids; its reference rules are public HTTPS and a catalog count.

BFL's FLUX 3 Image takes 1 to 10 reference images, each between 256 x 256 px and 16 MP, as a URL or base64. Sume's FLUX ids today are FLUX.2 pro and flex, not FLUX 3, and the Sume Image API page states different reference rules: public HTTPS URLs only, and a per-model input_references count from the catalog. It publishes no pixel bounds for references, so a reference outside BFL's range is a case you test, not one the docs settle.
BFL's figures are from its FLUX 3 Image page, and Sume's from the Sume Image API page and errors page, all read 2026-10-02.
What are BFL's reference limits for FLUX 3 Image?
The parameter table says images is a string or list of 1 to 10 references, URL or base64, 256 x 256 px to 16 MP each. aspect_ratio defaults to auto, which follows the first reference and falls back to 1:1 when there is none. The result comes back as a polling_url, and result.sample has to be downloaded within 1 hour.
| Rule | BFL FLUX 3 Image | Sume Image API |
|---|---|---|
| Count | 1 to 10 | 0 up to the catalog max (10 for most, 16 for GPT Image 2.5) |
| Form | URL or base64 | Public HTTPS URL |
| Pixel bounds | 256 x 256 px to 16 MP | Not published |
| Shape rule | aspect_ratio auto follows first reference | Use aspect_ratio auto on edits |
Does Sume serve FLUX 3 Image?
The OpenAPI model enum for POST /v1/images lists black-forest-labs/flux.2-pro and black-forest-labs/flux.2-flex as its FLUX ids; FLUX 3 is not in it. For FLUX 3 edits you would call BFL directly, with BFL's rules above.
What happens with a reference Sume cannot read?
Sume's docs say localhost, private-network and non-HTTPS URLs are rejected before submission. A public URL that Sume cannot fetch or mirror shows up as image_not_fetchable or input_media_unreachable, and the errors page says to check that the input is a public HTTPS image URL. The Sume docs show references only as public HTTPS image_url entries, so host the file instead of sending base64.
How do I stay safe on size?
Resize references to a size you have already seen succeed before you batch. Treat that as a habit, not a Sume rule: the Sume docs set no minimum or maximum.
Sources
Related posts
More in Comparisons
- Gemini batch create is not idempotent: two jobs vs Sume bulk replay
Gemini's docs say sending the same batch creation request twice creates two batch jobs. Sume bulk runs replay the old queue for the same Idempotency-Key.
- Flow charges 7 to 15 credits for Omni Flash; Sume bills dollars
Google Flow prices Gemini Omni Flash 720p at 7 to 15 credits by length. Sume bills the same model per second in dollars; clip costs from 3 to 10 seconds.
- Flow video edit costs 40 credits a try; Sume bills per second
Flow charges 40 credits per Gemini Omni video edit. Sume's video_url edit bills output seconds: $0.625 for 5 seconds, $1.25 for 10 at 720p.
- GPT-6 Sol vs Luna vs Astra: context, prices, cutoffs
GPT-6 Astra, Sol and Luna share a 1,050,000-token window but differ 100x in price. A table from OpenAI's own pages, plus which one Sume runs Formats on.
Written by Sume