Skip to content
Sample image made with FLUX.2 klein 4B
FLUX.2 Image Local Image input

FLUX.2 klein 4B

Pick FLUX.2 klein 4B when you want quick images from plain sentences and simple one-image edits. It is a step-distilled FLUX.2 model that renders in 4 steps and runs on your own GPU in Imference Desktop, not in the cloud. It favors speed over tiny details and ignores negative prompts.

$0

per image in the cloud

$0

on your own GPU

Good at

  • Fast drafts and iterations from short plain sentences.
  • One-image instruction edits with clear, targeted changes like color, background, or pose.
  • Reliable layout and subject placement when moving scenes or changing context.
  • Trades very fine texture detail for speed.

Running it

  • Local only: runs on your GPU in the free Imference Desktop app. No hosted endpoint.
  • Family: FLUX.2. Model code: flux2-klein-4b.
  • Accepts an optional single image for instruction edits. Text-only works too.
  • Use from code via the local REST API, and let agents pick it through the Imference MCP server.

Prompting it

  • Write plain sentences. Examples: "a foggy harbor at dawn", "portrait, soft rim light, 50mm".
  • For edits, attach one image and give a direct instruction: "make the jacket red", "place the subject on a beach".
  • Negative prompts are ignored by this step-distilled model.
  • Defaults are tuned: steps 4, guidance 1. You can set them in the API.

Use it

Two ways to run FLUX.2 klein 4B

Local · free

  1. Open Imference Desktop and stay on Local.
  2. In Local models, pick FLUX.2 klein 4B. It downloads once.
  3. Write a prompt and press Generate. It runs on your GPU.
Make an image with FLUX.2 klein 4B ("flux2-klein-4b"): a lighthouse on a cliff at dusk, cinematic light

Claude needs the MCP server.

Specs

The details

Price
$0 · 0 credits
$5 buys
0 images
Prompt style
Plain sentences
Runs
Local
Default steps
4
Default CFG
1
Reference images
Up to 1
Model code
flux2-klein-4b

Similar

You might also like

All 85 models

Try FLUX.2 klein 4B tonight