FLUX 3 Image
FLUX 3 Image is the image generation and editing model in Black Forest Labs' FLUX 3 family, released on 1 October 2026, ten weeks after BFL introduced FLUX 3 as a multimodal effort spanning video, images and robot actions. One model handles text-to-image and edits. It renders natively at 1K, 2K or 4K, takes up to 10 reference images, and accepts optional bounding boxes on a 0 to 1000 grid to place named elements or confine an edit to one region. Unlike FLUX.2, it is priced per image by resolution tier rather than per megapixel.
Anvisha Pai, Co-founder & CEO, Voyager
Black Forest Labs · released
›What we asked for
A self-portrait of FLUX 3 Image, Black Forest Labs' newest image model, as a Swiss-style typographic poster: a large red numeral '3' overlapping a black grid of thin rules, small aligned captions reading 'FLUX 3 IMAGE' and 'BLACK FOREST LABS', off-white paper, generous margins, precise layout.
blackforestlabs/flux-3/text-to-image
enable_prompt_expansion: false, aspect_ratio: 4:3, resolution: 1k, output_format: png
returned 1184 by 880 pixels
At a glance
- Price
- One flat price per image by tier on BFL's API: $0.041 at 768 square, $0.048 at 1K, $0.100 at 2K and $0.607 at 4K. fal's regular 1K price is the same $0.048.
- Native 2K and 4K
- Five tiers from 512 or 768 square up to 4K, about 16 megapixels, rendered at that size rather than upscaled. fal warns that 4K can take several minutes.
- Up to 10 references
- Edits and compositions take 1 to 10 reference images, referred to in the prompt by position, at one flat price however many you send.
- Bounding boxes
- On BFL's API you can name elements and place each in a box on a 0 to 1000 grid, then recolour, move, replace or remove them box by box while the rest of the picture stays unchanged.
- No negative prompt or seed
- BFL's API has neither field, and fal's endpoint has no seed either, so a result cannot be replayed. BFL advises describing what you want instead of what to leave out.
Other FLUX models: FLUX.2 Pro, FLUX.2 Flex, FLUX.2 Max. Compare all 4.
Anvisha's take
FLUX 3 Image is the best FLUX at text and layout I have run. The poster, the infographic and the chart's labels came back word for word with nothing added, and the woodblock print arrived unsigned, which none of the FLUX.2 tiers managed. Its misses are telling: the chart prints the right numbers on bars of the wrong heights, and the edit kept every pixel of the dripper, added the brass handle and never turned it green. It took 21 to 40 seconds for most pictures at 1K, about twice FLUX.2 Pro, and the edit took nearly two minutes. At $0.048 a 1K image it sits between FLUX.2 Pro and Max, and on these prompts I would take it over Max.
FLUX 3 Image real output: the same 12 prompts every model gets
Every image model on this site gets the same 12 prompts, so you can compare like with like. These are the pictures FLUX 3 Image returned, one try each, nothing re-rolled or retouched. Generated on . The 12 pictures took 21 to 112 seconds each and cost $0.288 in total.
Spelling, kerning and hierarchy of three lines of set text.
›What we asked for
A concert poster that reads 'MODA SUMMER SESSIONS' in large bold type, with the subtitle 'Friday 12 June, 8pm, Roundhouse' and a smaller line 'Tickets at the door'. Risograph style, two colours, tangerine and navy, grainy ink texture.
blackforestlabs/flux-3/text-to-image
enable_prompt_expansion: false, aspect_ratio: 3:4, resolution: 1k, output_format: png
returned 880 by 1184 pixels
Catalogue lighting, a clean seamless background and a believable contact shadow.
›What we asked for
Studio product photograph of a matte black ceramic pour-over coffee dripper on a seamless white background, soft top light, subtle contact shadow, centred, catalogue style.
blackforestlabs/flux-3/text-to-image
enable_prompt_expansion: false, aspect_ratio: 1:1, resolution: 1k, output_format: png
returned 1024 by 1024 pixels
A natural face, correct finger count and hands holding an object.
›What we asked for
Close-up portrait of a woman in her 40s with short grey hair, laughing, holding a red ceramic mug with both hands, natural window light, 85mm lens, shallow depth of field.
blackforestlabs/flux-3/text-to-image
enable_prompt_expansion: false, aspect_ratio: 3:4, resolution: 1k, output_format: png
returned 880 by 1184 pixels
A flat single-colour mark with a legible wordmark and no gradients or photo texture.
›What we asked for
A flat vector logo mark for a bicycle repair shop called 'Spoke & Chain': a single-colour icon combining a chain link and a wheel spoke, with the wordmark below in a geometric sans-serif, black on white.
blackforestlabs/flux-3/text-to-image
enable_prompt_expansion: false, aspect_ratio: 1:1, resolution: 1k, output_format: png
returned 1024 by 1024 pixels
Many distinct objects and people in one frame without merging or duplicating them.
›What we asked for
A crowded Saturday farmers market seen from slightly above: at least twelve stalls with striped awnings, shoppers carrying tote bags, a dog on a leash, crates of oranges, a musician with a guitar, bunting overhead, late morning light.
blackforestlabs/flux-3/text-to-image
enable_prompt_expansion: false, aspect_ratio: 4:3, resolution: 1k, output_format: png
returned 1184 by 880 pixels
A named historical style applied faithfully rather than a generic painterly look.
›What we asked for
A lighthouse on a rocky cliff during a storm, in the style of a Japanese ukiyo-e woodblock print, with visible wood grain texture, flat colour areas and a bokashi gradient sky.
blackforestlabs/flux-3/text-to-image
enable_prompt_expansion: false, aspect_ratio: 3:4, resolution: 1k, output_format: png
returned 880 by 1184 pixels
Six labelled bars at the stated heights, in order, with readable axes.
›What we asked for
A clean bar chart titled 'Monthly active users' showing six months, January through June, with the values 120, 150, 180, 170, 220 and 260, labelled axes, white background, one accent colour, no 3D effects.
blackforestlabs/flux-3/text-to-image
enable_prompt_expansion: false, aspect_ratio: 4:3, resolution: 1k, output_format: png
returned 1184 by 880 pixels
A 21:9 composition that leaves the left third clear for headline text.
›What we asked for
Website hero banner: a wide panorama of a sunlit modern co-working space with plants, the left third of the frame kept as clear wall space for text, warm neutral palette.
blackforestlabs/flux-3/text-to-image
enable_prompt_expansion: false, aspect_ratio: 21:9, resolution: 1k, output_format: png
returned 1568 by 672 pixels
Four numbered steps in the given order with their exact wording, plus an icon each.
›What we asked for
An infographic card titled 'How to brew pour-over' with four numbered steps in this order: 1. Rinse the filter, 2. Bloom for 30 seconds, 3. Pour in slow circles, 4. Wait 3 minutes. A simple icon beside each step, muted palette, clean sans-serif type.
blackforestlabs/flux-3/text-to-image
enable_prompt_expansion: false, aspect_ratio: 3:4, resolution: 1k, output_format: png
returned 880 by 1184 pixels
Appetising, menu-grade food with correct textures and no melted anatomy.
›What we asked for
Overhead photograph of a margherita pizza on a wooden board, fresh basil leaves, one slice pulled out with the cheese stretching, restaurant menu photography, natural light.
blackforestlabs/flux-3/text-to-image
enable_prompt_expansion: false, aspect_ratio: 1:1, resolution: 1k, output_format: png
returned 1024 by 1024 pixels
Exact counts and left-right placement of five named objects.
›What we asked for
A glass of water on a wooden table with exactly three ice cubes in it, a yellow pencil resting across the rim of the glass, a blue book to the left of the glass and a red apple to the right, photographed from table height.
blackforestlabs/flux-3/text-to-image
enable_prompt_expansion: false, aspect_ratio: 4:3, resolution: 1k, output_format: png
returned 1184 by 880 pixels
Change one object's colour and add a part while leaving background, light and framing untouched.
›What we asked for
Change the coffee dripper to a deep forest green and add a small brass handle on its right side. Keep the background, lighting and composition exactly as they are. (applied to the product on white result above)
blackforestlabs/flux-3/edit-image
enable_prompt_expansion: false, aspect_ratio: 1:1, resolution: 1k, output_format: png
returned 1024 by 1024 pixels
How the cost under each picture was worked out: fal pricing API at run time: $0.024 per megapixel, one 1 MP image at the 1k tier, flat whatever the number of references (fal.ai/models/blackforestlabs/flux-3/text-to-image, read 2026-10-08); fal's 50% launch promotion through October 8, 2026, regular $0.048; fal does not return the charge.
Compare FLUX 3 Image
FLUX 3 Image pricing
| Where | Unit | Price | Source |
|---|---|---|---|
| BFL API | 1K image, about 1 megapixel One flat price per image by tier, for generation and edits alike: $0.041 at 768 by 768, $0.100 at 2K (about 4 megapixels), $0.607 at 4K (about 16 megapixels). The submission response returns each request's cost. | $0.048 | BFL pricing, read |
| BFL API | 2K image, about 4 megapixels | $0.10 | BFL pricing, read |
| BFL API | 4K image, about 16 megapixels | $0.607 | BFL pricing, read |
| fal | 1K image, 1 megapixel Regular price, flat whatever the number of references. fal ran a 50% launch promotion, $0.024 at 1K, which it said ends on 8 October 2026; our samples ran on the last day of it and record the rate that was live. fal does not publish its 2K and 4K figures on the model page. | $0.048 | fal model page, read |
Is FLUX 3 Image free?
No free tier. Black Forest Labs sells FLUX 3 Image per image through its API, with a playground for testing, and partners such as fal and Replicate charge per request.
- •BFL API: credits at $0.01 each, charged per image by resolution tier.
- •fal and Replicate: pay per request; neither lists a free allowance for this model.
- •Third-party creative apps may bundle FLUX models with their own sign-up credits.
BFL pricing, read
Where to use FLUX 3 Image
- BFL API · POST /v1/flux-3-image for generation and edits, with bounding boxes and search grounding.
- fal · blackforestlabs/flux-3/text-to-image and blackforestlabs/flux-3/edit-image; the endpoints used for the samples below.
- Replicate · Hosted under black-forest-labs, generation and editing in one model.
Voyager
Run FLUX 3 Image inside Voyager
Voyager is an agent for creative work that runs every model in this reference, FLUX 3 Image included. Brief it, and it picks the model, makes the thing and hands back something you can edit. Create an account to get started.
What we noticed running FLUX 3 Image
- •All three lines correct, in the case written, and nothing added. A two-colour risograph layout with the date and venue on a navy band, closer to a designed poster than the FLUX.2 tiers' plain stacks. See the text in an image picture.
- •Four steps in order with exactly the wording given, each with a matching icon, and no invented instructions underneath. See the infographic with steps picture.
- •Title, months, values and axis are all right, but the bars do not match their own labels: 120 stands at about 65 on the axis and 170 at about 130. Check bar heights, not just the numbers. See the chart picture.
- •Half an edit. The handle turned brass and the rest of the picture is unchanged, the left 60 percent of the frame identical to the source pixel for pixel, but the body stayed black instead of forest green. It took 112 seconds, the slowest request in the run. See the edit an existing image picture.
- •No painted signature, seals or title cartouche on the woodblock print. FLUX.2 Pro, Flex and Max all added one to the same prompt. See the specific art style picture.
- •The left third is plain wall, with no text or button painted in. The 21:9 request came back at 1568 by 672. See the wide aspect ratio picture.
- •Book left, apple right, pencil across the rim and three ice cubes in the glass, as asked. See the hard composition picture.
- •Wordmark spelled right with no tagline added. The icon reads more as a chain link than as a wheel spoke. See the logo mark picture.
- •Most pictures took 21 to 40 seconds at 1K, median 26, against about 12 for FLUX.2 Pro on the same prompts. The 21:9 banner took 66 seconds and the edit 112.
- •The 1k tier is about one megapixel in each shape: 1024 by 1024 square, 880 by 1184 at 3:4, 1184 by 880 at 4:3. You choose a ratio and a tier, not exact pixels.
FLUX 3 Image limits and API parameters
| Limit | Value |
|---|---|
| Output size | Fixed tiers rather than exact pixels: 768 by 768, about 1, 4 or 16 megapixels at 1K, 2K or 4K, in fifteen aspect ratios from 21:9 to 9:21. BFL pricing, read |
| Reference images | Up to 10 per request on fal, each at least 256 pixels per side; larger images are downscaled to at most 4 megapixels. fal edit model page, read |
| Negative prompts | Not supported. BFL's guidance is to describe the wanted result, an empty promenade rather than no crowds. |
| Transparent background | No transparency option on either API; fal outputs JPEG or PNG without an alpha setting. fal API schema, read |
›Every setting, for developers
fal (blackforestlabs/flux-3/text-to-image and /edit-image)
fal API schema, read
| Parameter | Type | Values | Default |
|---|---|---|---|
| prompt | string | Any text Required. | |
| aspect_ratio | enum | auto, 21:9, 2:1, 16:9, 3:2, 7:5, 4:3, 5:4, 1:1, 4:5, 3:4, 5:7, 2:3, 9:16, 1:2 Auto follows the first reference image when editing. | auto |
| resolution | enum | 512sq, 768sq, 1k, 2k, 4k 4k can take several minutes. | 1k |
| image_urls | list of URLs | 1 to 10 images, each at least 256 pixels per side Edit endpoint only, where it is required. Larger inputs are downscaled to at most 4 megapixels. | |
| output_format | enum | jpeg, png | jpeg |
| safety_tolerance | integer | 0 to 4, 0 is strictest | 2 |
| enable_prompt_expansion | boolean | true, false Rewrites the prompt while keeping its intent and the roles of the references. | false |
| version | string | A pinned model version | latest |
BFL API (flux-3-image)
BFL API reference, read
| Parameter | Type | Values | Default |
|---|---|---|---|
| images | URL or base64, or a list | 1 to 10 reference images | |
| aspect_ratio | enum | auto, 21:9, 2:1, 16:9, 3:2, 7:5, 4:3, 5:4, 1:1, 4:5, 3:4, 5:7, 2:3, 9:16, 1:2, 9:21 | auto |
| resolution | enum | 768sq, 1k, 1.5k, 2k, 4k | 1k |
| grounding | boolean | true, false Lets the model use web and image search to inform the prompt. | true |
| safety_tolerance | integer | 0 to 4, 0 is strictest | 2 |
What FLUX 3 Image will not draw
Black Forest Labs licenses the hosted model for commercial use, and fal marks the endpoints commercial. Both BFL and fal run a moderation dial, safety_tolerance from 0, strictest, to 4, defaulting to 2. The usual prohibitions apply: sexual content involving minors, non-consensual intimate imagery and deceptive depictions of real people. There is no adult mode.
BFL API reference, read
Alternatives to FLUX 3 Image
Frequently asked questions
What is FLUX 3 Image?
FLUX 3 Image is Black Forest Labs' image generation and editing model, released on 1 October 2026 as the image part of the FLUX 3 family. It renders natively up to 4K, takes up to 10 reference images and can place elements with bounding boxes.
How much does FLUX 3 Image cost?
On BFL's API it is a flat price per image by size: $0.041 at 768 square, $0.048 at 1K, $0.100 at 2K and $0.607 at 4K, read on 8 October 2026. fal's regular 1K price is also $0.048, after a 50% launch promotion that ran to 8 October.
Is FLUX 3 Image better than FLUX.2?
It adds native 2K and 4K, box-placed layouts and edits that leave the rest of the picture alone, at a price between FLUX.2 Pro and Max for a 1K image. Compare the twelve prompts on this page with the FLUX.2 pages, or see the head-to-head with FLUX.2 Max.
Does FLUX 3 Image have an API?
Yes. BFL's endpoint is /v1/flux-3-image. fal hosts it as blackforestlabs/flux-3/text-to-image and blackforestlabs/flux-3/edit-image, and Replicate as black-forest-labs/flux-3-image.
How this page is made
Prices, limits and settings are read from the linked vendor and host pages on the dates shown. Every picture was generated by us through fal with the request shown under it, and the original files are kept unedited. A generation that failed is shown as a failure. Model and vendor names are used to identify the products; no vendor imagery appears on this page.
