FLUX 3 Image Klein

Sogni runs FLUX 3 Image Klein, the fast 5.6B open-weights version of Black Forest Labs' FLUX 3 Image, on the Sogni Supernet: on one RTX 5090 it renders a 1024×1024 image in about 1.9 s and edits a photo in about 3.6 s (see Speed). One model covers three jobs:

• Text-to-image from a prompt alone.
• Image-to-image from a start image, with a strength that sets how closely the result follows it.
• Edits and compositions guided by one to ten reference images.

References are optional: unlike Qwen Image Edit and Krea 2 Identity Edit, it renders from a prompt alone when you attach none. The output is always the width and height you request. Example renders, a comparison with Sogni's other open image models, and the full price table are on the FLUX 3 Image Klein model page.

For side-by-side renders and timings against the Krea 2, Qwen and Z-Image models on Sogni, see FLUX 3 Image Klein Speed Test on the Sogni blog.

#Speed

Measured on one NVIDIA RTX 5090 with each model already loaded, at each model's default steps. Time in the app also includes the queue and transfers.

The speed comes partly from size: Black Forest Labs puts Klein at 5.6 billion parameters, smaller than Krea 2 (12 billion, per Krea) and Qwen Image (20 billion, per Qwen), so it trades some ceiling for speed. Black Forest Labs got a lot out of a small model, but a bigger one was still ahead in some of Sogni's tests: Krea 2 Identity Edit drew a clearer tiny catapult, and Qwen Image Edit 2511 Lightning made the more toy-like plastic figurine.

#Workflow ID

#Limits

Renders larger than 2048×2048 pixels, including the 4K presets, run only on 32 GB GPUs and take longer; see Pricing for measured times. 4K is for text-to-image: a one-photo edit runs at the photo's own size, and when a 4-megapixel photo was edited at 3840×2160 the result came out blotchy, and a 2048×2048 photo edited at 4096×4096 lost the edit.

Sogni Socket rejects a request with more than ten reference images before it is priced or queued, rather than dropping the extras. FLUX 3 Image Klein is distilled to sample at guidance 1, which is why there is no guidance control and why a negative prompt changes nothing: put everything you want into the prompt itself.

#Personal LoRAs

Sogni Unlimited members can import their own FLUX 3 Image Klein LoRAs from Civitai or Hugging Face in My LoRAs in Sogni Web, or through the personal LoRA API, and render with them. Klein imports form their own family: a LoRA imported for Krea 2 or Qwen Image Edit does not apply to Klein, and a Klein LoRA does not apply to them. The strength starts at the value the author used in the Civitai version's example images, or 1 when there is none. Sogni has no catalog LoRAs for Klein yet.

#Reference images

• Order matters. Refer to references by position in the prompt: "the first image", "the second image", and so on. In the SDK the first entry of contextImages is the first image.
• Reference sizes follow Black Forest Labs. A single reference keeps its own shape and is used at up to about 4 megapixels, never enlarged. With two or more, each is shrunk to at most 1 megapixel. The output size is the width and height you send, not the size of a reference.
• Edit one photo at its own shape. Send the photo's own size (multiples of 16, up to about 4 megapixels) so everything outside the edit stays put; a different shape, such as a square from a wide photo, makes the model recompose the scene. Sogni Web and the creative tools choose that size for you.
• A start image or references, not both. The reference-image workflows have no start-image input, so send startingImage or contextImages in one render. Sogni Web stops a render that carries both and asks you to remove one.
• More references cost more. Each one adds processing time, which the estimate includes (see Pricing).

#Layout boxes

FLUX 3 Image Klein reads layout boxes written into the prompt:

1. Tag each element in the description with an id in angle brackets, for example a large hot-air balloon <obj_1> floats over the valley.
2. End the prompt with a JSON list of boxes. Each box has the id, a bbox of [top, left, bottom, right] on a 0–1000 grid measured from the top-left corner, and a short desc.

... a large hot-air balloon <obj_1> floats over the valley, while a smaller blue hot-air balloon <obj_2> hangs farther back on the right ... [{"id": "obj_1", "bbox": [80, 60, 480, 380], "desc": "A large red and yellow striped hot-air balloon"}, {"id": "obj_2", "bbox": [200, 620, 420, 800], "desc": "A small blue hot-air balloon with a tiny brown basket"}]

In an edit, a box can also carry "from": "target" with a tgt_bbox to place a new element in the result, or "from": "ref_image_0" with a src_bbox and tgt_bbox to say where an element sits in the first reference and where it goes. Boxes guide placement and scale; they are not hard masks. The model page shows complete prompts that use both forms.

In Sogni Web you don't have to write boxes by hand: attach one photo, turn on Edit by box, draw up to three boxes and choose Add object, Move or resize or Remove object for each. Sogni writes them into the prompt in this format.

#Prompting

• In Sogni Web, short prompts are expanded for you. A FLUX 3 Image Klein render runs the Prompt Enhancer first: a short prompt becomes a full scene caption, and an edit instruction is folded into a caption of the photo. Prompts already written as a full caption, and Edit by box renders, are sent as they are. Turn off Auto, under the Enhance button, to send your words exactly as typed.
• Write a full scene caption. Black Forest Labs trained the model on long, detailed captions, and its model card notes that short prompts produce generic output. Describe the subject, framing, setting, lighting, and style in plain sentences rather than a list of keywords.
• Put exact text in quotes. Quote the words you want in the image and say where they go and how they look, for example 'THE LOW TIDES' very large at the top, in black wood type. Ending with "nothing else written" keeps stray lettering out. Large display type is the most reliable; Black Forest Labs notes that rendered text can be inaccurate, so check small print at full size.
• Say what each reference contributes. Name in the prompt what to take from each image, such as a face, a product, a room, or a palette, and how they come together.
• Say what stays the same. For an edit, add what must not change, such as the background, the light, and the framing.
• Phrase exclusions positively. Ask for "a clean, empty desk" rather than "no clutter"; a negative prompt has no effect at guidance 1.
• Trade steps for speed. 16 steps is the default; fewer steps, down to 4, render faster with less fine detail and cost proportionally less.

#Pricing

FLUX 3 Image Klein is a standard open model on Sogni, not a Premium Spark or partner model, so it is covered by Sogni Unlimited plans under fair use and can also run on pay-as-you-go Spark. Values below use the standard 1 Spark = $0.005 peg, at the default 16 steps.

Text-to-image starts at 1.80 Spark ($0.009) per image at the default 16 steps, for any size up to about 1.5 megapixels; larger images scale with pixels (1536×1536 is 2.71 Spark), and fewer steps cost proportionally less (a 4-step 1024×1024 render is 0.45 Spark). A one-photo edit grows faster than its pixels because the photo is edited at its own size: on an RTX 5090 it took 3.6 s at 1 megapixel, 8.3 s at 2 and 22.1 s at 4. 4K renders run on 32 GB GPUs and take longer: on an RTX 5090, a 3840×2160 render took 27 s and a 4096×4096 render took 94 s. Each reference image also adds render time: on an RTX 5090 at 1024×1024, renders with 7, 8, 9 and 10 references took 22.5, 26.9, 31.6 and 36.6 s. Sogni shows the estimate before the job runs; call projects.estimateCost() for a live estimate in your own app.

#Where you can use it

• Sogni Web: pick it under Next-Gen Diffusion in the model picker, or choose it from Image Editing on Home to start an edit.
• API and SDKs: model ID flux3_image_klein_sd_int8_convrot (example below).
• Creative tools: model: "flux-3-klein" on generate_image and edit_image (see Creative tools).

#JavaScript SDK example

import { SogniClient } from '@sogni-ai/sogni-client';
import fs from 'node:fs';

const sogni = await SogniClient.createInstance({ appId: 'my-app', network: 'fast' });
await sogni.account.login('username', 'password');

const project = await sogni.projects.create({
  type: 'image',
  network: 'fast',
  modelId: 'flux3_image_klein_sd_int8_convrot',
  positivePrompt:
    'Product photo of the ceramic mug from the first image on the oak counter of the kitchen ' +
    'in the second image, morning window light from the left, shallow depth of field',
  numberOfMedia: 1,
  sizePreset: 'custom',
  width: 1024,
  height: 1024,
  steps: 16,
  // Optional: leave contextImages out for plain text-to-image.
  contextImages: [fs.readFileSync('./mug.png'), fs.readFileSync('./kitchen.jpg')]
});

const [url] = await project.waitForCompletion();

For image-to-image, pass startingImage and startingImageStrength (0–1) instead of contextImages.

#Creative tools

In hosted chat completions and creative-agent workflows, set model: "flux-3-klein" on generate_image for text-to-image or a start-image render (sourceImageIndex with starting_image_strength), and on edit_image for edits guided by up to ten reference images.