The Freiburg lab's site now carries an "Introducing FLUX 3 Image" banner with API access at 50 percent off through October 8, and a free demo is live at flux-tools.bfl.ai. FLUX 3 itself was unveiled in late July as a multimodal foundation model spanning still images, 20-second video clips with synchronized audio, and action prediction — with the Image product promised "in the coming weeks." This is that release: the still-image line going wide while FLUX 3 Video and FLUX 3 Action remain in a gated Early Access program.

What it does#

The headline capability is multi-step editing without collateral damage: revise one region of an image across several passes, BFL claims, without changing the rest of the picture. Users compose scenes with bounding boxes, feed in up to ten reference images, and get outputs up to 4K. The model covers text-to-image, image-to-image, text rendering, and photorealism — the full checklist the current generation of image models is judged on.

Split-view showing a portrait whose background was swapped while the subject stayed identical
Region-precise editing: same subject, new background — AI-generated illustration, AI Frontier Post

How BFL is selling it#

The launch-week discount is the land grab: half-price API access through October 8 to pull developers in fast. Behind it sits a two-track commercial strategy. Companies can license commercial weights to run and fine-tune the model on their own infrastructure — the enterprise track — while an open-weight version is expected in the coming weeks, which would put FLUX 3 Image in the hands of anyone with the hardware to run it. That combination of cheap API now, licensed weights for business, and open weights soon is BFL's bid to own the image-editing layer across every deployment shape at once.

Designer adjusting a high-resolution display showing a cityscape with a pixel-level selection active
Pixel-level selection on a 4K canvas — AI-generated illustration, AI Frontier Post

Ideogram got there first — by days#

Shortly before BFL's launch, Ideogram announced its own editing-focused model, version 4.5, which is also set to ship as open weights soon. Two labs converging on the same idea — precise, multi-turn editing, then open weights — within days of each other is not a coincidence; it is where the frontier of image generation has moved. The generation-quality race is largely settled. The editing race is just starting.

Why it matters#

For production work — ad creatives, product shots, film previz — the question was never whether a model could make a pretty picture. It was whether a director could change the third object from the left without the other eleven shifting. Regional, multi-turn editing is the control layer that decides real adoption, and it is now the explicit battleground: BFL with FLUX 3 Image, Ideogram with 4.5, both promising open weights within weeks. Whoever ships the most controllable editor with the most permissive weights wins the next wave of image tooling. The next few weeks will tell.