Breaking

Black Forest Labs Introduces FLUX 3 Image With 4K Output and Multi-Reference Editing

October 1, 2026 at 15:00 EDT

Foundation Models
Media Generation

Black Forest Labs introduced FLUX 3 Image, a still-image generation and editing system aimed at precise commercial workflows, with promised pixel-level control, bounding-box layout direction, output up to 4K and composition from up to 10 reference images. The announcement moves the image branch of FLUX 3 from roadmap item to product after the company first described FLUX 3 as a unified multimodal foundation model for images, video, audio and action prediction in July.

Oct. 1, 2026 · Black Forest Labs

FLUX 3 Image aims at pixel-level control

The new image model promises precise multi-turn edits, bounding-box layouts, up to 4K output and composition from as many as 10 references.

reference images for one composition

maximum image output named in the launch

1,026

image-prompt pairs in earlier KontextBench work

The reference jump is the workflow signal

1 refbasic guidance

10 refsidentity, product, style and scene cues

The advertised ceiling supports more deliberate composition, where separate inputs can carry different roles instead of forcing everything into one prompt.

How the workflow changes

Compose

Use prompt, references and bounding boxes.

Edit

Change one part over multiple turns.

Preserve

Keep untouched pixels stable.

Why creators care

The pitch maps to real production pain: product consistency, exact layouts, readable assets and fewer accidental changes during revision.

What remains unclear

Public materials do not yet provide independent image benchmarks, detailed endpoint docs, per-image pricing or open-weight terms for the image capability.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.