FLUX 3 Image Launches: Pixel-Precise Editing, Bounding Boxes, 4K Output

The feature list

• Pixel-precise multi-turn editing — change one element across editing rounds while the rest of the frame stays visually consistent
• Bounding-box layout control — place elements with integer coordinates on a normalized 0-to-1000 grid
• Up to 10 reference images per generation — style, character, and composition references combined in one prompt
• Up to 4K output resolution — BFL's samples include a 5,456 x 3,072 frame
• Text rendering — short English lettering on labels, packaging, and posters
• Search grounding enabled by default to verify visual details

How the bounding-box system works

FLUX 3 Image maps every canvas to a normalized 0-to-1000 grid measured from the top-left corner, independent of aspect ratio or pixel dimensions. Each element gets an ID, a natural-language description, and a box in [top, left, bottom, right] order. A layout combines the element table with a scene-level caption, and IDs plus coordinates stay editable between turns — so a layout can be stored as project state and revised element by element by an agent or a designer.

Access and pricing

FLUX 3 Image is available now through the BFL Playground and API, with 50% off API pricing through October 8, 2026 as a launch promotion. Companies can also license commercial weights to run and fine-tune the model on their own infrastructure. An open-weight release is expected "in the coming weeks," with no calendar date announced — consistent with BFL's pattern on earlier FLUX generations, where hosted access preceded open weights by weeks to months.

The competitive picture

FLUX 3 Image lands in an open-weights race: Ideogram announced its editing-focused Ideogram 4.5 shortly before the FLUX 3 launch, also slated to ship as open weights. The two take different routes to the same goal — FLUX 3 Image uses explicit JSON element tables with persistent IDs and coordinates, suited to programmatic, agent-driven pipelines; Ideogram 4.5 works from a source image plus natural-language instruction and a mask, friendlier to designers who want no layout-authoring overhead.

“Introducing FLUX 3 Image. Control every pixel. Make precise multi-turn edits without changing any other pixel.”

Who this is for

The positioning is squarely professional: designers, marketers, and content teams doing magazine covers, product composites, ad creatives, and collages where element placement and spacing matter as much as style. Multi-reference composition plus editable layouts moves AI image generation from one-shot prompting toward a repeatable design workflow — provided teams pressure-test the edit-consistency claim on real assets first.