The news
Black Forest Labs has made FLUX 3 Image, the image model in its FLUX 3 family, available to developers, with native generation at up to 4K resolution and editing guided by on-canvas boxes, according to the company's model page. FLUX 3 Image generates and edits images in a single model, the company says.
Black Forest Labs says the model renders natively at 2K and 4K, up to 5,456 by 3,072 pixels, or 16.8 megapixels. It can combine up to 10 reference images in one generation and edit content inside individual bounding boxes without changing the rest of the image, the company says. The model page describes FLUX 3 Image as natively trained to understand image layout and composition, which Black Forest Labs pitches for workflows where AI agents plan a layout.
Runware, an AI inference platform that hosts the model, lists support for text-to-image, image-to-image, multi-reference generation, local editing and text rendering, including text in multiple languages. Runware says generation and editing use the same model and the same request format, so teams do not need separate tools for each step.
Runware lists four resolution tiers with launch-week prices, which it describes as 50% off, available through October 8: $0.021 per image at 0.75K, $0.024 at 1K, $0.05 at 2K and $0.304 at 4K. Black Forest Labs also offers access through its own API dashboard and playground, according to its model page, which did not list prices.
On licensing, the Black Forest Labs model page points to open weights on Hugging Face and GitHub under a non-commercial license, and a separate commercial weights license for enterprise-scale deployment.
The numbers
- Maximum native resolution
- 5,456 x 3,072 pixels (16.8 MP)
- Reference images per generation
- Up to 10
- Runware launch-week price, 4K
- $0.304 per image
- Runware launch-week price, 0.75K
- $0.021 per image
- Black Forest Labs valuation (per DataNorth)
- $3.25 billion
Why CEOs should care
For marketing and creative leaders, FLUX 3 Image is another serious option outside the biggest platform companies. Box-level editing and up to 10 reference images target the work agencies and brand teams do every day: keeping a product, logo placement or character consistent across many variations. Ask for a pilot on your own brand assets and compare edit accuracy and text rendering against the generator you use now.
CFOs should look closely at per-image economics. Runware's launch-week prices run from about two cents at the smallest tier to about 30 cents at 4K, and those are discounted rates that end October 8. Model the cost at the resolution you actually need; for social and web assets, lower tiers may cover most of the volume.
Legal and procurement teams should focus on the licensing split. The open weights carry a non-commercial license, so commercial self-hosting needs the separate commercial weights license. For companies with strict data rules, the option to run the model on their own infrastructure under that license can matter more than price. Confirm who owns outputs, what indemnity is offered and what restrictions apply before rolling the model out.
The bigger picture
FLUX 3 Image is part of a wider FLUX 3 family that Black Forest Labs announced on July 23, 2026, covering video with synchronized audio, image generation and action prediction for robotics, according to DataNorth. The video and action models entered early access first, with the image model following, DataNorth reported.
DataNorth reported that Black Forest Labs is valued at $3.25 billion, has raised more than $450 million from investors including a16z, Nvidia (NVDA), Salesforce Ventures and Adobe Ventures, and employs about 100 people across Freiburg, Germany, and San Francisco. That backing, including from Adobe's venture arm, shows how independent model makers are competing with and supplying the large creative software and cloud platforms.
What’s next
Watch for standard pricing on Runware after October 8, published prices on Black Forest Labs' own API, and more third-party platforms adding FLUX 3 Image, which would make it easier for enterprises to test it inside tools they already use.
What “Fact-checked” means
Fact-checking means testing a story’s facts against the evidence before it is published. This story went through at least two separate checks before this version was published.
- What we checked
- Its names, figures, dates, job titles, quotes and who said what were checked against the story’s sources, including its main source where it could be opened. The headline was checked for accuracy and overstatement.
- How
- A first check reviewed the whole story. If it passed, a second, skeptical check went back to the sources to look for mistakes in the most important facts. If a check flagged the story, it was edited to fix the problems found, and a separate re-check then reviewed the whole story again.
- Who
- The checks are made by our newsroom, as steps kept separate from the writing, under rules set by our editor, Hussein Mukhtar. A story the checks still flag is held for the editor, who decides whether it is fixed, published or dropped.
- If something is wrong
- “Fact-checked” does not mean error-free. If a material error is found after publication, we correct the story and add a note saying what changed. Report an error
