Black Forest Labs' newest image model. Create native 4K images with sharp text, place every element exactly where you want it, and edit one detail while the rest of the scene stays in place.
Model
Image (Optional)
Click to upload an image
PNG, JPG, JPEG, or WEBP (max 10MB)
Prompt
Aspect Ratio
Resolution
Experience the power of AI image generation
Flux 3 Image is the image generation and editing model in Black Forest Labs' FLUX 3 family, released on October 1, 2026 as the successor to FLUX.2. It turns a written idea, up to ten reference images, or a layout of boxes into a finished image at native 4K, and it edits your existing pictures with the same model.
What sets Flux 3 Image apart is control. You can tell it where every element goes on a 0 to 1000 grid, and when you edit, you name what changes and what stays, so the rest of the scene keeps its layout and light. It researches unfamiliar subjects with web and image search before it draws, and it renders small text and fine textures at native 4K without an upscaling pass.
Flux 3 ImageFLUX 3 is one model family with separate models for video, images, and robots. This page covers the one that makes and edits pictures.
Black Forest Labs unveils one model family for images, video, audio, and robot actions.
Text and image to video with synchronized sound, through the Black Forest Labs API.
A 7B open-weights model that turns visual understanding into robot control.
Text to image, 10 references, box layouts, and targeted edits up to native 4K.
Generate at 1K, 2K, or 4K in your browser, no API key needed.
Open the generatorDates from Black Forest Labs' announcements. Flux 3 Image weights are licensed to companies for self-hosting, not released openly. FLUX 3 Video and FLUX 3 Action are separate models and do not run in this image generator.
Street scenes, product shots, infographics, portraits, and posters, all generated with this model. Hover any image to see the exact prompt behind it.
Four upgrades over FLUX.2 that give you control over resolution, text, references, and edits.
Every detail drawn, nothing upscaled. Flux 3 Image renders up to 5456 Ă— 3072 pixels (16.8 megapixels) in a single pass. Faces in a crowd, fabric weave, and small print stay sharp because the model draws them at full size instead of enlarging a smaller image.
Flux 3 ImageWords that read right at any size. Put the words in quotes and Flux 3 Image sets them as written, from a billboard headline to a price tag in the corner. It handles Latin, Chinese, and Japanese scripts in the same image, so menus, packaging, and bilingual posters come out ready to use.
Your person, your product, your place, in one shot. Upload up to ten images and refer to each one in your prompt. Flux 3 Image keeps faces, products, and locations recognizable while it builds a new scene around them, so a portrait, a product photo, and a street become one campaign image.
Flux 3 ImageChange two things. Keep everything else. Name the elements you want to change and the ones to keep, each with its box. Flux 3 Image redraws the changed boxes and keeps the rest of the scene in place, with the same layout, light, and objects, so you can swap furniture, recolor clothes, or replace a sign without starting over.
You draw the layout, the model fills it in. Name each element in your prompt and give it a box on a 0 to 1000 grid, and it composes the image around your layout at any aspect ratio. Tap a row to see its box.
Flux 3 ImageElement table from the prompt
Flux 3 Image succeeds FLUX.2 as Black Forest Labs' flagship image model. Same prompt below, both rendered at 2 megapixels.
Same prompt, two models: A street food chalkboard menu reading "TACOS AL PASTOR $3", "HORCHATA $2" and "OPEN LATE", a taquero slicing pork from a glowing trompo behind it, warm night light, candid photo.
| NewestFlux 3 Image | FLUX.2 [max] | |
|---|---|---|
 | ![FLUX.2 [max] result for the taco stand chalkboard prompt](https://cdn.pixmira.ai/bananaai/flux-3-image/compare-flux-2-max.webp) | |
| Model family | FLUX 3, October 2026 | FLUX.2, 2025 |
| Maximum output | 4K, 16.8 MP (5456 Ă— 3072) | 4 MP (2048 Ă— 2048) |
| Reference images | Up to 10 | Up to 8 |
| Layout control | Bounding boxes on a 0 to 1000 grid | Text and JSON prompts |
| Editing | Box by box, the rest kept in place | Instruction-based edits |
| Aspect ratios | 15, from 21:9 to 9:21 | 9, from 16:9 to 9:16 |
| Best for | 4K detail, exact layouts, targeted edits | Fast high-quality images up to 4 MP |
Swipe the table to compare both models
Specs from Black Forest Labs' FLUX 3 Image documentation and the Replicate model pages, October 2026.
Choose Flux 3 Image whenever resolution, placement, or a clean edit matters: it renders four times the pixels of FLUX.2 [max], takes two more reference images, and edits box by box. To compare it with GPT Image 2.5 or Nano Banana 2.1, switch models in the same generator.
Compare all modelsPick the size you need. Every option is rendered natively, with no upscaling, and costs a fixed number of credits per image.
| 1K | 2K | Native 4K4K | |
|---|---|---|---|
| Resolution | 1 megapixel | 4 megapixels | 16.8 megapixels |
| Credits per image | 6 | 12 | 70 |
| Best for | Social posts, drafts, and quick tests | Websites, ads, and presentations | Print, large banners, and close crops |
Swipe the table to see all three sizes
Credits per image on PixMira. At 4K, Flux 3 Image outputs up to 5456 Ă— 3072 pixels (16:9).
From idea to finished image in four steps with the Flux 3 Image generator at the top of this page.
It is already selected in the generator at the top of this page.
Upload up to 10 images of the people, products, places, or styles you want it to use. Skip this step to create from text alone.
Write the subject, setting, light, and style. Put any words you want in the image inside quotes, and add boxes when you want exact placement.
Choose an aspect ratio and 1K, 2K, or 4K, then generate. To refine, upload the result and describe only what should change.
Name each part of the image and put exact text in quotes. Flux 3 Image follows the structure piece by piece.
A red vintage scooter parked outside a pastel blue bakery in Lisbon, morning sun casting long shadows, a shop sign reading "PADARIA", shot on 35mm film.
Quick answers about Flux 3 Image and how to use it on PixMira.
Your next poster, product shot, or room makeover is one sentence away.
Start creating