FLUX 2 AI Image Generator – The FLUX.2 Model Family Online

FLUX 2, written FLUX.2 by Black Forest Labs, is the image model family released before FLUX 3. Every FLUX 2 model does text to image and editing with reference images. The hosted versions differ in size and focus: [pro] is the production default, [flex] is tuned for typography, and [max] is the largest and most detailed.

Black Forest Labs lists photorealism, multi-reference editing, readable text and exact color control with hex codes as the family's main strengths. Smaller FLUX 2 versions also exist ([klein] and [dev]) with open weights for local use. flux-3image runs the three hosted models: [pro], [flex] and [max].

Model	What it's for	References	Credits at 1K	Page
FLUX.2 [pro]	Everyday photoreal images, products, careful edits	1–8	3	FLUX 2 Pro
FLUX.2 [flex]	Posters, labels, menus and other text-heavy designs	1–8	5	FLUX 2 Flex
FLUX.2 [max]	Maximum detail and the hardest edits	1–8	6	FLUX 2 Max

Start with FLUX.2 [pro]. Black Forest Labs calls it the recommended default model of the family, and it costs the fewest credits of the three. Move to [flex] when the image is mostly lettering, and to [max] when [pro] keeps missing a fine detail. The same prompt works on all three FLUX 2 models, so switching takes one click.

Photorealism is where FLUX 2 puts most of its effort. This kitchen shot came from a two-sentence prompt describing the chef, the dish, the copper pans and the light. Skin, fabric folds, steam and the shallow depth of field read like a documentary photograph, and the second cook in the background stays softly out of focus.

Describe the scene the way a photographer would plan it: who is in it and what they're doing, where it happens, where the light comes from and how close the camera is. Short, concrete sentences beat long lists of style words.

For wide scenes, FLUX 2 handles repeating patterns well. The tea terraces above follow the curve of the hills row by row, and the road winds through them. The single red scooter we asked for is there too, small enough to give the scene scale.

Wide formats suit this kind of image. Pick 16:9 or 21:9 for banners and headers, and name one small focal point (a scooter, a hiker, a boat) so the eye has somewhere to land.

Top-down product arrangements are a common FLUX 2 AI job for shops and blogs. The prompt listed five objects and the linen surface. FLUX.2 [pro] spaced them evenly, kept them unbranded as asked, and added a soft window shadow across the cloth.

For exact brand colors, add hex codes next to the objects, as in "a notebook in #E9DCC9". FLUX 2 maps each code to the object it sits next to. The FLUX 2 Pro page shows a hex-color edit step by step.

FLUX 2 is not free on flux-3image. Each FLUX.2 [pro], [flex] or [max] image costs credits, because every request is a paid call to the model's API. Accounts created with Google sign-in get a few free credits, enough for a first FLUX 2 test before you buy anything. FLUX.2 [pro] costs the fewest credits per image.

Some sites offer a free "FLUX 2" that is actually one of the small open-weights versions running on their own hardware. Those are real FLUX models, but they are not [pro], [flex] or [max].

1. Open the model menu and choose a FLUX 2 model. This page starts on FLUX.2 [pro].
2. Write your prompt. Start with the subject, then add setting, light and camera. Put hex codes next to the objects they color.
3. Add references if you're editing. Upload the image to change, or as many as eight images to combine, and say what each one contributes.
4. Choose 1K or 2K and an aspect ratio between 21:9 and 9:16.
5. Generate. Change one thing in the prompt per re-run, so you can tell what made the difference.

FLUX 3 Image is the newer model, released on October 1, 2026. It takes 1 to 10 references and outputs 4K. FLUX 2 still makes sense when you want a lower credit cost per image, hex-code color work or text-heavy designs on [flex]. A common workflow is to explore with FLUX 2, then render the final version with FLUX 3 Image on the home page.