FLUX 3 Image is live from Black Forest Labs as of Oct. 1, 2026: a generate-and-edit image model that lays out scenes with bounding boxes, accepts up to 10 reference images, outputs native 2K and 4K, and routes create and edit traffic through a single API surface. Early coverage from The Decoder and the company’s own bfl.ai product page emphasize multi-step edits that leave untouched regions alone—an agent-friendly design for layout-heavy creative work.
Part of our Tech, AI and Business coverage. Browse more in Science, Health and Culture.
Bounding-box layout is the headline feature
Instead of hoping a free-form prompt places every object correctly, FLUX 3 Image lets users or agents draw boxes on a 0–1000 canvas and describe each element. A global scene caption ties the pieces together; an element table (id, bbox, description) travels with the prompt. Black Forest Labs says boxes the user draws reach the model verbatim, while an optional upsampler can densify captions without rewriting those coordinates.
That workflow suits posters, editorial spreads, collages, and crowded scenes where typography and figures must lock to precise regions. Developers can also skip hand-drawn boxes and let an LLM propose the element table from one line plus an aspect ratio—then nudge only the boxes that look wrong before generating.
References, 4K output, and one generate/edit endpoint
FLUX 3 Image accepts up to 10 reference images, tokenized in order as ref_image_0 and beyond, so streetwear composites, product mockups, or character-consistent scenes can pull garments, props, and faces from separate files. Native high-resolution renders aim to keep small type, fabric texture, and facial detail intact without a separate upscaler stage.
Editing is framed as the same model path: re-describe a box, replace a subject, recolor a garment, or move an element while locking everything else. The Decoder notes text-to-image, image-to-image, text rendering, and photorealism in the same family, with a free demo on the BFL site. API pricing is discounted 50 percent through Oct. 8, 2026, according to launch coverage, giving studios a short window to benchmark cost against rivals.
Commercial weights and the open-weight roadmap
Companies can license commercial weights to fine-tune and serve on private infrastructure—important for brands that cannot send every creative asset to a public endpoint. Black Forest Labs has also signaled an open-weight release in the coming weeks, paralleling other labs racing to ship editable generative stacks. Ideogram’s editing-focused 4.5 line was announced in the same news cycle, sharpening competition on controllable layout.
For product readers comparing consumer try-on and generative fashion tools, our recent piece on the ChatGPT Virtual Changing Room shows how mainstream chat apps are chasing apparel visualization even as specialist image labs push pixel-level control. Security-minded teams evaluating frontier multimodal APIs may also weigh rollout patterns similar to those in our Gemini 4 Argon coverage.
What creators and agents should test first
Practical evaluation checklists include: multi-box poster layouts with readable headlines; multi-turn recolors that truly freeze backgrounds; ten-reference fashion boards; and 4K crops of fine text. Agent builders should verify that planned element tables survive the upsampler and that edit records remain auditable for brand review.
FLUX 3 Image will not replace every diffusion workflow overnight, but bounding-box composition plus locked-region edits address the most common production complaint—unwanted global drift after a local change. Studios that already pipeline LLM layout planners into image APIs are the most likely early adopters before the introductory API discount ends on Oct. 8. Broader policy context for generative tools continues to evolve; see our report on Bill Gates pressing Congress on AI regulation for the compliance conversation surrounding rapid model releases.
AlphaSignal and other industry digests have already flagged FLUX 3 Image as a notable October launch for controllable generation, underscoring how quickly layout APIs are becoming table stakes for creative agents. Teams should document box conventions early so marketing, legal, and engineering share one vocabulary for locked versus editable regions.
Sources
- The Decoder: Black Forest Labs launches Flux 3 Image with multi-step editing that leaves the rest of your picture alone
- Black Forest Labs: FLUX 3 Image: Maximum control over every pixel
Further Reading
- Long COVID Dopamine Study Finds Loss of Brain Nerve Endings
- Dolly Parton Day Brings Parades, Pink Lights, Books
Read next in Artificial Intelligence
Artificial IntelligenceOpenAI Says Coding Agents Now Finish 46% of Two-to-Four-Hour Research Tasks on Their Own
Artificial IntelligenceMeta Says AI Agents Now Handle More Than 60% of Its Ads-Ranking Model Upgrades
Artificial IntelligenceCalifornia Panel Urges Emergency Power to Shut Down Dangerous AI Systems
Most read this week
- Finland Confirms Strike on Russian Warship in Gulf of Finland; Ukraine’s SBU Claims Drone Attack
- Nvidia Earnings, Walmart and Fed Minutes Headline a Busy Week for Wall Street: What to Watch Nov. 16-20
- Strait of Hormuz Insurance Rates Explained: What War-Risk Cover Costs in 2026
- DSA-Backed Candidates Won 106 Races on Nov. 3, Including 8 House Seats: Full Results
- Rare Earth Truce Explained: What the Jan. 10 Deadline Means
Latest
- Congress Nears a Deal to Ban Chinese Cars From the U.S., With Changes to Spare Mercedes-Benz
- Phil Eng Agrees to a Dec. 7 Council Session, Ending the Blue Hill Avenue Subpoena Standoff
- Massport Is Testing Autonomous Wheelchairs and 4 Cleaning Robots at Logan
- MBTA’s Top Two Police Officials Will Retire Nov. 30; Union Warns of ‘Precipice of Chaos’
- MBTA Weekday Ridership Hit 1,041,000 in October, First Time Above 1 Million Since 2020
Get The Boston Report Briefing
Our free weekly email with the most important stories from Boston and the world. Unsubscribe anytime.
Free. We never sell your address. Privacy policy