Flux 3 Image Launches with Multi-Step Editing and 4K Image Generation

Black Forest Labs released Flux 3 Image, the image generation and editing component of the Flux 3 model family, on October 2, 2026. A free demo was also made available alongside the company’s announcement. The model combines text-to-image generation, editing existing images, text generation, and photorealistic results.

What features does Flux 3 Image offer?

One of the model’s headline capabilities is its aim to preserve areas that users do not want changed while making successive edits to an image. This feature is described by the company; the source does not include an independent performance evaluation. As a result, outputs will need to be checked at every step, especially for work requiring precise visual control.

Users can use bounding boxes to set up a scene and add up to 10 reference images to the generation process. Output resolution can reach 4K. These options make it possible not only to describe the overall direction of an image, but also to incorporate the position of specific elements or reference material into the workflow.

Flux 3 Image is not limited to generating images from scratch. Support for text-to-image and image-to-image generation also lets users transform an existing image based on a new prompt. Text generation is another listed capability, for work that requires lettering within an image. However, the source does not specify which languages these features support or provide any measurements of text accuracy.

Access, discount, and open-weight plans

The commercial options announced by Black Forest Labs include API access and commercial weights that can be licensed for companies to run on their own infrastructure and fine-tune. API access is 50% off through October 8. The source does not specify the discount’s starting price or the cost of different usage scenarios.

An open-weight version is expected in the coming weeks, so it was not available at the time of the announcement. Teams planning to run the model on their own infrastructure or fine-tune it will need to watch for the release of the weights and further details about the licensing terms. The free demo was announced as a separate way to try the model.

Impact on architecture and visualization workflows

For architects, interior designers, and visualization teams, the most notable potential use is the promise of editing an existing scene image step by step. For example, the goal is to change only selected parts of an image while preserving the rest. Reference images and bounding boxes may also be useful for teams looking to define a scene more precisely. This does not mean the model integrates directly with architectural software; the source provides no information about a specific 3D software plugin or compatibility.

Offices considering adding Flux 3 Image to their existing pipeline should first test it with their own types of imagery in the free demo, then assess the API or licensing options. For commercial projects in particular, it is important not to build a workflow around the model until output review, terms of use, and total costs are clear. The open-weight release is another development worth tracking for teams planning to run the model on local infrastructure.