Black Forest Labs image model with bounding-box layout control, reference-image input and native 4K output.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
FLUX 3 Image is a text-to-image and image-to-image model from Black Forest Labs. Its main feature is layout control: users or agents place elements on a canvas with bounding boxes and per-box descriptions, and can edit a finished image one box at a time while leaving the rest unchanged. The model is described as natively trained to understand image layout and composition.
The vendor's pages list text-to-image generation, editing, use of up to 10 reference images, and rendering at full resolution (a sample is 5456 × 3072 pixels, and the page mentions native 4K). The pricing page lists it as a 5.6B-parameter model that runs on a single consumer GPU, alongside a smaller FLUX 3 Image Klein variant.
It is accessed through Black Forest Labs' pay-as-you-go API with no subscriptions or seat fees. A commercial weights license is offered to companies that want to fine-tune and deploy it on their own infrastructure, by contacting the vendor. Public weights are not available at present. A YouTube creator noted that the vendor has said weights will be opened later, but that the same was promised for FLUX 3 Video and has not happened.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
A closed-source image model with bounding-box composition control, text-to-image, up to 10 reference images and natural language editing. Weights are promised to be open later, but the creator notes Flux 3 video was promised the same and still isn't open.
Teased unified multimodal model for images, video with audio (up to 20s), and robotics action prediction. Creator is skeptical of the quality versus Seedance; preview only with early access request.
Multimodal foundation model that learns jointly from images, video and audio. It generates and edits images and video, and aims at a world model with action prediction. The creator says its videos look similar to Seedance or Kling.
Closest neighbours by meaning, not just by category.
Model · Black Forest Labs
5.6B-parameter open-weights model from Black Forest Labs for image editing, text-to-image and multi-reference composition.
Model · Black Forest Labs
Black Forest Labs' video generation model in the FLUX 3 family, with native audio and clips up to 20 seconds.
Model · Black Forest Labs
9B-parameter open-weights image generation and editing model from Black Forest Labs, built for low-latency use.
Model · Black Forest Labs
7B open-weights world action model from Black Forest Labs that jointly predicts video and actions for robot control.
Model · OpenMOSS
Open-weights 5B text-to-image model family from OpenMOSS with native JSON bounding-box layout control and text rendering.
Model · PrismML
PrismML's 1-bit/ternary compressed Flux 2 Klein text-to-image model (~1 GB) for offline on-phone generation.