Is FLUX 3 Image open source? Commercial weights or the BFL API

Not in the way the question usually means. Black Forest Labs' FLUX 3 Image page says the model is available under a commercial weights licence for companies running image generation at scale, and through BFL's own API. There is no free download with a permissive licence on that page: weights come with a paid, self-hosted commercial licence.

Sume's image catalog on main lists FLUX.2 pro and FLUX.2 flex, not FLUX 3 Image, so you cannot call FLUX 3 through Sume today. All facts here come from BFL's pages and Sume's docs, read on 2026-10-03; the launch itself is reported by The Neuron's October 1 digest, not by BFL.

What does BFL's page say about weights and licences?

The FLUX 3 Image page offers two routes: a commercial weights licence for self-hosting, and the BFL API. It links four legal documents: the Non-Commercial License, the Self-Hosted Commercial License, the Developer Terms of Service and the FLUX API Service Terms. The licensing page describes the self-hosted tiers as Builder, Platform, Professional and Enterprise, all with fine-tuning and LoRA rights, and says "License FLUX model weights. Fine-tune on your data, self-host on your infrastructure."

The pages I read do not say that FLUX 3 Image weights are downloadable under a non-commercial licence, the way FLUX.2 dev is. If you need that answer, check the Hugging Face and GitHub listings and the licence file itself before you plan around it.

What does FLUX 3 Image do that people want?

BFL's page lists several edits at once through bounding boxes while leaving the rest of the picture unchanged, up to 10 image references, and native 2K and 4K output with a maximum of 5456 by 3072 pixels. The Neuron's digest describes the same region-based editing on a 0 to 1000 canvas and says the playground needs a sign-in. Those are the reasons teams ask about weights: region edits on private product shots are something a company may want on its own servers.

Which should you use: weights, BFL API or Sume?

For regional edits on Sume, the documented tool is a mask. The Image API takes an optional public HTTPS mask_url on ChatGPT Image 2.5 edits, and the docs say a request that sets a parameter the chosen model does not list is rejected with 400 unsupported_parameter. That is a different edit method from BFL's boxes: you draw a mask, not coordinates. The post on pixel-exact local edits compares the two in more detail.

• Weights licence: you run the GPU, you own the uptime, and you pay BFL for the licence. It suits high volume and private data.
• BFL API: you call FLUX 3 Image directly and follow BFL's service terms. It is the only listed way to call the model without self-hosting.
• Sume's Image API: you get FLUX.2 pro or flex, plus other families, under one key and one wallet. It does not offer FLUX 3 Image, bounding-box edits or custom weights.

How do you check what Sume lists today?

The catalog changes. This script prints every FLUX id Sume's image catalog lists, so you can see whether a FLUX 3 row has been added since this post:

import os
import requests

r = requests.get(
    "https://api.sume.com/v1/images/models",
    headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"},
    timeout=30,
)
r.raise_for_status()
for m in r.json()["data"]:
    if "flux" in m["id"]:
        print(m["id"], m.get("supported_parameters", {}).get("input_references"))

What if a FLUX 3 row appears?

If a FLUX 3 row appears, its supported_parameters will say which of the edit fields it accepts. Until then, treat FLUX 3 Image as a BFL-side decision: pick the API or a weights licence, and use Sume for the models it does list.