by Black Forest Labs·Released 15 October 2026
FLUX.3 Image Klein is a model you can download and run yourself. It makes images from a written description and edits them with reference images, holds a character or subject steady across a set of pictures, borrows a style from a reference and draws legible text. Within the FLUX.3 Image line it is the smaller, quicker model.
You pay for what each request uses. No subscription, no minimum.
One image, the ordinary use
Every rate
| Meter | ↑ you send | ↓ comes back |
|---|---|---|
| Picture | 3,00 DA |
per picture returned
One company runs this model. OpenDunes sends every request to it.
| Provider | Per image | Latency | Uptime |
|---|---|---|---|
| Black Forest Labs | 3,00 DA | … | … |
You pay the price of the provider that served your request. Prices read 15:20 · speed over the last 30 minutes · uptime over 3 days.
One HTTP call. Any language that can send JSON can call it.
import os, requests
BASE = "https://opendunes.com/v1"
KEY = os.environ["OPENDUNES_API_KEY"]
H = {"Authorization": f"Bearer {KEY}"}
resp = requests.post(f"{BASE}/images", headers=H, json={
"model": "black-forest-labs/flux-3-image-klein",
"prompt": "A clay tagine on a blue zellige table, morning light",
})
print(resp.json())Runs this request once on your balance.
Standard OpenAI fields, grouped by what they control. Anything not listed is accepted and ignored.
Shaping the reply
How long, how varied, where to stop.
seedAbout 3,00 DA per image.
No. You pay per use from your balance, with no subscription. Try it in Chat at the same price.
FLUX.3 Image Klein is a model you can download and run yourself. It makes images from a written description and edits them with reference images, holds a character or subject steady across a set of pictures, borrows a style from a reference and draws legible text. Within the FLUX.3 Image line it is the smaller, quicker model.
FLUX 3 Image costs 12,98 DA more for the same image.
Deposit dinars with your Edahabia or CIB card. Every model, FLUX 3 Image Klein included, is paid from that balance. No foreign card, no conversion.
Black Forest Labs names no languages for FLUX 3 Image Klein. Arabic, French and Darja: not stated.
Yes. Up to 10 reference pictures in one request.
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 and 9:21. Saved as JPEG and PNG.
Black Forest Labs. Released 15 October 2026.
FLUX.3 Video makes video with matching sound from a written description, an image used as the opening and closing frame, or an existing clip you want to continue. Clips run 5 to 20 seconds, from HD up to 4K, in shapes up to 21:9. It suits controlled scenes, bringing a still picture to life and extending footage.
Send a clip and a written instruction to FLUX Video Edit and get the same clip changed: objects or characters added, removed or swapped, a new setting, rewritten on-screen text, new colors or materials, or translated speech with matching lip sync. Length, shape and soundtrack stay as filmed. The clip must be shorter than 15 seconds and under about 50 MB, and anything sharper than HD is reduced to HD.
FLUX Video Upscale takes one finished video and enlarges it 1.5 to 3 times, leaving the length unchanged, with an optional written instruction to guide the result. Choose the precise mode to stay faithful to the original, or the creative mode to let the model add detail. Use it to sharpen footage for a bigger screen. It creates nothing new: for fresh clips use FLUX.3 Video, and to change what happens in a clip, FLUX Video Edit.
#8 for photo editing · Black Forest Labs' speed tier of FLUX.2, a 4B model made for high-throughput pipelines that need many images quickly while holding the image quality its maker describes as excellent
From what its maker and two benchmark sites publish, and where it stands on OpenDunes. Read 7 Nov 2026.
Up to 10 references
Takes up to that many pictures to guide the result.
4K
Makes pictures up to 4K.
Transparent background No.
SVG No.
Several pictures per request No.
Returns more than one picture per request.
Vertical 9:16
Same result twice
The same seed gives the same result.
Keep one character recognisable across a series of edited scenes
Up to 10 references
Confine an edit to a drawn region so the rest of the picture stays as it was
Run image generation on your own graphics card from downloaded weights