What you can do with FLUX 3 Image Klein here
FLUX 3 Image Klein is the 5.6B open-weights model in the FLUX 3 family. Running it yourself means downloading about 52 GB of weights, including its prompt upsampler, and setting up a GPU; here it runs in the box above.
- Text to image. Describe the scene and get one image at 1K, 2K, or 4K, in fourteen aspect ratios or Auto.
- Edit with an instruction. Attach a photo and write the change. Black Forest Labs built Klein to change what you name and keep the rest of the picture.
- Combine up to ten references. Add up to ten images and refer to them in order: the first image, the second image.
- Lettering. Put a short title or label in quotation marks; Black Forest Labs' own Klein samples include posters, infographics, and seed packets.


Price per image
Every run shows its price on the prompt box before you start. Text prompts and edits cost the same, and the number of reference images does not change the price.
| Resolution | FLUX 3 Image Klein | FLUX 3 Image |
|---|---|---|
| 1K (about 1 MP) | 4 credits | 18 credits |
| 2K (about 4 MP) | 6 credits | 72 credits |
| 4K (about 16 MP) | 14 credits | 245 credits |
Credits come with a Lite, Pro, or Ultra plan; see the pricing page. If a generation fails, its credits come back automatically.
Klein or FLUX 3 Image?
Both models are in the model menu above, with the same controls.
- Pick Klein to iterate fast and cheaply: drafts, variations, batches of ideas, and edits where you want one thing changed. In our first runs a 1K image came back in about five seconds.
- Pick FLUX 3 Image when you want the larger flagship model, or need bounding-box layouts: this workbench reads boxes written in the prompt for FLUX 3 Image, but does not send them to Klein yet.
Black Forest Labs has not published a quality comparison between the two, so the honest test is your own prompt on both. Need motion instead of a still? FLUX 3 Max turns a prompt or a start frame into a short clip with sound. Background on the model, its license, and running it locally is in FLUX 3 Image Klein: what it is, price & VRAM.
How to use FLUX 3 Image Klein
- Sign in with Google.
- Check that the model menu on the prompt box says FLUX 3 Image Klein.
- Write the scene, or attach images and write the change you want.
- Pick an aspect ratio and a resolution; the credit price updates on the prompt box.
- Run it. Finished images land in your history.
Prompt tips for Klein
- Write it like a caption. Klein is trained on detailed captions: say the framing, the subject, the background, and the light.
- Short prompts still work. This workbench turns a short prompt into a detailed caption before rendering, which is how Black Forest Labs recommends running Klein.
- For edits, name the change. "Turn the summer street into winter after light snowfall" works better than describing the whole new picture.
- Quote exact words. Put lettering in quotation marks and say where it goes.
FAQ about FLUX 3 Image Klein
What is FLUX 3 Image Klein?
A 5.6B-parameter FLUX 3 image model with open weights, released by Black Forest Labs on October 15, 2026. It handles text to image, editing, and up to ten references in one model.
Can I use FLUX 3 Image Klein online without a GPU?
Yes. This page runs it in the browser: sign in, write a prompt, and the image renders on our side. No download, no setup.
How much does FLUX 3 Image Klein cost here?
4 credits at 1K, 6 at 2K, and 14 at 4K per image, for text prompts and edits alike.
Can FLUX 3 Image Klein edit my photos?
Yes. Attach a photo and write the change as an instruction. Klein is built to change what you name and keep the rest of the image.
How many reference images can I add?
Up to ten per run. Refer to them in order in the prompt: the first image, the second image.
Is FLUX 3 Image Klein free here?
No. Runs use credits from a plan, and there is no free tier. Failed runs refund their credits automatically.
Is this the official FLUX 3 Image Klein site?
No. This is an independent workbench, not affiliated with, endorsed by, or sponsored by Black Forest Labs. FLUX is their trademark, used here only to identify the model we provide access to.
