FLUX 3 vs FLUX 2: which one you can call on Sume today

For images you can call through Sume today, the choice is between FLUX.2 Pro and FLUX.2 Flex, because Sume's image catalog has no FLUX 3 id. BFL's release notes say FLUX 3 is a phased rollout that begins with video, and the API entry they list is FLUX 3 Video, not an image endpoint.

BFL's side is from its release notes and Sume's from the Image API docs and catalog code, all read 2026-09-29.

Does FLUX 3 replace FLUX.2 for stills?

Not according to the pages read. BFL describes FLUX 3 as covering video, keyframe control, image synthesis and editing, and says the rollout starts with video. The release notes I read list no separate FLUX 3 image endpoint, so I cannot tell you when FLUX 3 stills arrive or how they compare with FLUX.2.

How do I call FLUX.2 on Sume?

Put a FLUX.2 id in model. The Image API docs show black-forest-labs/flux.2-pro in their request example; swap in black-forest-labs/flux.2-flex to use Flex.

curl -X POST https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "black-forest-labs/flux.2-flex", "prompt": "a shop sign that reads OPEN, brushed steel"}'

Which of the two should I start with?

Run the same prompt through both ids and compare the results and the pricing lines on GET /v1/images/models. The catalog is the place to see what each id accepts before you build around it.

What differs between the two FLUX.2 ids?

Sume's docs do not describe a quality gap between black-forest-labs/flux.2-pro and black-forest-labs/flux.2-flex, so this post does not either. What the catalog does give you is a per-model capability list and a pricing line, and those are the reliable way to see a difference: fetch both records and compare supported_parameters and cost_usd.

Per-image cost for each id is worked out in FLUX.2 API pricing.

Why not wait for FLUX 3 stills?

Waiting has no date attached. BFL's notes give the rollout order, video first, and the pages read state no schedule for image output. A project with a deadline should ship on what the catalog lists now and re-check the list later.

When you do re-check, run your own prompts through the new id next to the old one and keep the outputs; a model name alone says little about how your prompts will render.

How do I compare the two ids fairly?

Fix the prompt, aspect ratio and number of images, and change only the model value. Save both outputs with the request you sent and the usage.cost you were billed, then judge them against the brief you actually have. One prompt is an anecdote, so use a small set that includes the hardest case you expect, such as a sign with text or a crowded scene, and repeat it before you decide which id becomes your default.