FLUX 3 Image

FLUX 3 Image is a prompt-conditioned image generation and editing model from Black Forest Labs, the creator behind the flux family. It unifies generation and editing through a single endpoint, letting the prompt itself decide whether the model produces a new image or modifies an existing one rather than relying on a separate mode switch. This unified design is positioned around fine-grained visual control rather than raw generation alone, with the model framed as giving "maximum control over every pixel."

In practice, FLUX 3 Image stands out for its layout-aware prompting, accepting a JSON list of bounding boxes appended directly to the prompt to place elements such as headlines, faces, or grid panels without any extra parameter. It also performs pixel-exact local edits in one request, allowing multiple regions to be recolored, replaced, moved, resized, or removed while the rest of the image stays untouched, and it can incorporate up to ten named reference images for restyling or multi-image composition. Output scales from 768-square previews up to 4K (about 16 megapixels) across fifteen aspect ratios, making it well suited to workflows that range from quick ideation to high-resolution production assets where precise spatial control matters.

Vercel AI Gatewaybfl/flux-3-imageflux

Open provider page Provider docs

Quick Info

Provider: Vercel AI Gateway
Model key: bfl/flux-3-image
Release date: Oct 5, 2026
Last updated: Oct 5, 2026
Input modalities
Output modalities
Capabilities: Base catalog fields only

Limits

Output tokens: 0 tokens
Context window: 0 tokens

Latest news about FLUX 3 Image

Vercel AI Gateway

CoverageBenchmark

Kingy AI's October 1, 2026 launch analysis confirms FLUX 3 Image's GA via the BFL API and browser Playground, with hosted access at POST /v1/flux-3-image. Supported tasks span text-to-image, instruction-based editing, multi-reference composition, and layout-controlled generation, with up to 10 reference images supplied as URLs or base64, minimum 256×256 inputs, and a 20 MB base64 payload cap. The piece documents output tiers of 768sq, 1k, 1.5k, 2k, and 4K (default 1k) and 15 explicit aspect ratios from 21:9 to 9:21 plus auto. Commercial weights are available through BFL with a public open-weight release promised in the coming weeks. Kingy AI notes it has not run paid generations or matched hands-on benchmarks, so quality comparisons remain vendor-supplied.

Vercel AI Gateway

CoverageRelease Notes

Black Forest Labs released FLUX 3 Image on October 1, 2026, combining generation and editing through a single endpoint where the prompt alone decides the mode. The model supports bounding-box layout coordinates, pixel-exact local edits, and up to ten reference images named by prompt position. Output resolution scales from 768sq previews up to about 16 megapixels at 4K across 15 aspect ratios, giving developers fine-grained layout and reference control in one API call. Per-image pricing tiers are $0.041 at 768sq, $0.048 at 1k, $0.100 at 2k, and $0.607 at 4K, with the endpoint exposed at the FLUX 3 Image API. Recolor, replace, move, resize, and remove operations can be batched into a single request without a mode parameter. BFL also documents companion pages for the box format and bounding-box tutorial to help developers integrate layout-driven image generation.