Black Forest Labs/flux-3-image/Released Oct 2026

FLUX 3 Image

Text-to-image and multi-reference editing, 4K

Commercial use
Image
FLUX 3 Image

About FLUX 3 Image

FLUX 3 Image generates images from a text prompt, edits an existing image from a written instruction, and combines up to ten reference images into one output. Output ranges from 768 x 768 drafts to roughly 16 megapixels at 4K, across 15 aspect ratios from 21:9 to 9:21.

Edits name what to change and leave the rest of the image untouched: recolor, replace, move or remove an element, or restyle the whole frame. With several references, each image is cited by its position ("image 1", "image 2") and can supply a subject, product, setting or style. For exact placement, a JSON list of bounding boxes appended to the prompt positions each element on a 0 to 1000 grid, both for new layouts and for targeted edits. Optional grounding lets the model search the web and images so real places, products and events are rendered accurately; without it, generation is faster and uses the prompt alone.

See what this model can do

A handful of real prompts and their outputs.

Macro photo of a garden spider at the center of an orb web strung with hundreds of dew drops, backlit by a rising sun.
Macro
Dew-Covered Web
"Hundreds of distinct refracting droplets on hair-thin silk, held sharp across the frame at 2K."
A white lattice pavilion with a wave-shaped canopy glowing at dusk and mirrored in a reflecting pool.
Architecture
Parametric Pavilion
"Clean repeating lattice geometry and a near-perfect mirror reflection under blue-hour light."
Two sumo wrestlers locked together in a dusty clay ring during morning practice as younger wrestlers watch.
Documentary
Morning Practice
"Grainy film-style reportage with shafts of dusty light and believable body contact."
A giant manta ray glides through open blue water while a free diver descends behind it.
Underwater
Diver and Manta
"Conveys scale and depth with sun rays, suspended particles and a crisp manta in the foreground."
The Milky Way arches over dead camelthorn trees standing on a cracked white clay pan with red dunes behind.
Astro Landscape
Milky Way over the Pan
"Long-exposure night photography look with pin-sharp stars and light-painted foreground trees."
A Moroccan riad courtyard with an octagonal zellige fountain, potted orange trees and carved cedar balconies.
Interiors
Riad Courtyard
"Dense tile patterns, carved woodwork and stucco arches rendered with consistent perspective."
An abandoned subway station overgrown by birch trees and ferns, with a deer drinking from a puddle between the rails.
Concept Art
Reclaimed Station
"Cinematic matte-painting composition with volumetric light through a collapsed ceiling."
An ice hockey goalie in red stretches out to catch a puck in his glove as ice spray flies in front of the net.
Sports
Glove Save
"Frozen high-shutter action with ice spray, correct goalie gear and no stray lettering."
Snow monkeys soaking in a steaming hot spring as snow falls, a mother cradling her baby behind an older monkey.
Wildlife
Hot Spring Macaques
"Snowflakes settling on wet fur, rising steam and soft winter light rendered at 2K."

Ready to build with FLUX 3 Image?

Try FLUX 3 Image in the Workbench to prompt it, compare outputs, and iterate on prompts without writing any code. When you're ready to ship, call the same model from our API and build your own apps on top of it.

curl -sSf -X POST https://hub.oxen.ai/api/ai/images/edit \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer $OXEN_API_KEY" \
    -d '{
  "model": "flux-3-image",
  "prompt": "A cat sitting on a chair",
  "aspect_ratio": "auto",
  "resolution": "1k",
  "grounding": false,
  "safety_tolerance": 2
}'

Request parameters

Every field FLUX 3 Image accepts in the request body. See the API reference for response formats.

POSThttps://hub.oxen.ai/api/ai/images/edit
FieldTypeRequiredDefaultDescription
promptstringYes
—
Prompt for the generated image. Use @Image1, @Image2, etc. to reference the input images.
input_imagearray<string>
nullable
No
—
Input images to edit or use as reference, up to 10. Reference them in the prompt as @Image1, @Image2, etc.
aspect_ratiostringNo"auto"
auto21:92:116:93:27:54:35:41:14:53:45:72:39:161:29:21
resolutionstringNo"1k"
768sq1k2k4k
groundingbooleanNofalseSearch the web and images before generating, so real places, products and events come out right. Slower than generating from the prompt alone.
safety_toleranceintegerNo2Safety filter strictness, 0 is strictest.
01234