Choose Auto, TokenLab Verified, or Official for each request, with prices shown up front. See what's new

Black Forest Labs: FLUX.2 [klein] 4B Base

FLUX.2 [klein] 4B Base is a compact undistilled image generation and editing model with an exceptional quality-to-size ratio. It is well suited for local deployment, efficient fine-tuning, and custom pipelines that require flexibility on limited hardware.
Compare models
flux-2-klein-4b-base
AvailableBlack Forest LabsImageSync
Price
From $0.0019
Modalities
Image

About FLUX.2 [klein] 4B Base

FLUX.2 [klein] 4B Base is the undistilled version of Black Forest Labs' smallest FLUX.2 model, with 4 billion parameters for text-to-image and multi-reference editing. Unlike the step-distilled klein 4B, it keeps the full training signal, which gives more varied outputs and a better starting point for fine-tuning, at the cost of needing many more sampling steps.

Where it works well

  • It is undistilled, so it keeps the complete training signal and is the intended base for LoRA training and custom post-training.
  • Black Forest Labs reports higher output diversity than the distilled klein models, useful when you want many different takes on one prompt.
  • At 4 billion parameters it fits in roughly 13GB of VRAM, within reach of consumer GPUs for local experiments.
  • Weights use Apache 2.0, the most permissive license in the FLUX.2 family.

When to choose another model

  • Its default sampling uses about fifty steps, so it is slower per image than the step-distilled klein 4B.
  • Text inside images can come out distorted, and the maker warns that prompt adherence depends on phrasing.
  • The larger FLUX.2 [dev] and 9B models handle complex scenes and reference blends with more capacity.

Getting started

  1. Create API key

    Create a key in Console, then use it with every model on the platform.

  2. Send your first request

    Copy the example for your language and run it against the endpoint.

    Image editingText to imageTokenLab endpoint
    POST/v1/images/generations
    curl -X POST "https://api.tokenlab.sh/v1/images/generations" \
      -H "Authorization: Bearer sk-xxx" \
      -H "Content-Type: application/json" \
      -d '{
      "operation": "image-to-image",
      "model": "flux-2-klein-4b-base",
      "prompt": "A minimalist product photo of matte black headphones on a soft blue background.",
      "image_url": "https://example.com/source.png"
    }'

Parameters

Image editing

ParameterValues and limitsDefaultNotes
cfg_scale≥ 1, ≤ 20Not providedUse multiples of 0.01.
guidance_scale≥ 1, ≤ 20Not providedUse multiples of 0.01.
height≥ 128, ≤ 20481024Use an integer. Use multiples of 16.
imageNot providedNot providedAlias of the combined images input. See input_limits.images for the total input image count.
image_urlNot providedNot providedAlias of the combined images input. See input_limits.images for the total input image count.
image_urlsNot providedNot providedAlias of the combined images input. See input_limits.images for the total input image count.
imagesInputs: ≥ 1, ≤ 4Not providedLimits apply when this input is supplied. Image aliases are combined into one input image list; the limit applies to the combined list.
n11Use an integer.
negative_promptNot providedNot providedMinimum length: 1 characters. Maximum length: 10000 characters.
operationRequiredNot providedNot providedNot provided
output_formatjpeg, png, webpNot providedDefault encoding is shared by these equivalent values: jpg, jpeg.
promptNot providedNot providedMinimum length: 1 characters. Maximum length: 10000 characters.
seed≥ 0, ≤ 9007199254740991Not providedUse an integer.
sizeNot provided1024x1024Use WIDTHxHEIGHT in pixels. The width and height constraints also apply to size.
steps≥ 1, ≤ 5028Use an integer.
width≥ 128, ≤ 20481024Use an integer. Use multiples of 16.

Text to image

ParameterValues and limitsDefaultNotes
cfg_scale≥ 1, ≤ 20Not providedUse multiples of 0.01.
guidance_scale≥ 1, ≤ 20Not providedUse multiples of 0.01.
height≥ 128, ≤ 20481024Use an integer. Use multiples of 16.
imagesNot providedNot providedText-to-image does not accept source images. Do not send this field for that operation. Use one of this model's declared image operations for source-image requests.
n11Use an integer.
negative_promptNot providedNot providedMinimum length: 1 characters. Maximum length: 10000 characters.
operationRequiredNot providedNot providedNot provided
output_formatjpeg, png, webpNot providedDefault encoding is shared by these equivalent values: jpg, jpeg.
promptNot providedNot providedMinimum length: 1 characters. Maximum length: 10000 characters.
seed≥ 0, ≤ 9007199254740991Not providedUse an integer.
sizeNot provided1024x1024Use WIDTHxHEIGHT in pixels. The width and height constraints also apply to size.
steps≥ 1, ≤ 5028Use an integer.
width≥ 128, ≤ 20481024Use an integer. Use multiples of 16.

Fields outside this contract are rejected.

Pricing

The Verified price applies to Verified, which costs less on most models. The Official price is the model maker's published price and applies to the more reliable Official route. Auto bills the route that completes the request.

Image editing

per request
Official price
$0.0019
Verified price
—
Discount
—

Text to image

per request
Official price
$0.0019
Verified price
—
Discount
—

Usage & activity

Success rate is the share of requests that completed. Latency is how long a full response takes; P95 means 95% of requests finished within that time.

Usage & availability

Last 24 hours
Requests
Success rate
P95 latency
Total tokens
Model performance
Metrics appear once privacy and data volume thresholds are met.

Data is based on aggregate user requests, excluding status checks.

Open in Console

Open FLUX.2 [klein] 4B Base in Console with a prompt ready to edit or send.

Help me create an image with flux-2-klein-4b-base using image-edit at /v1/images/edits. Show the result, cost, and any limits I should know.

Use cases

Best for
  • Text to Image
  • Fine-tuning a house style

    Train a LoRA on a few hundred brand images, starting from a base that has not been compressed by distillation.

  • Generating varied candidates

    Produce many different interpretations of one brief for a designer to pick from, benefiting from the wider spread of outputs.

  • Research on small image models

    Use it as the reference point when measuring what distillation or a new training recipe changes.

  • Light editing pipelines

    Run text-to-image and reference-based edits through one small model inside a custom workflow.

Prompt examples

A watercolor map of a fictional harbor town, muted teal and ochre, hand-lettered feel, no legible text.

Four different logo directions for a bakery called Crumb, flat vector style, one color each.

Take the sofa from the reference photo and show it in a sunlit Scandinavian living room.

FAQ

What does Base mean in FLUX.2 [klein] 4B Base?

Base means undistilled. The checkpoint keeps the full training signal rather than being compressed into a few-step sampler, so it is more flexible for customization and gives more diverse outputs, but it is not tuned for the fastest inference.

How is it different from FLUX.2 [klein] 4B?

The plain 4B is step-distilled for speed and can return an image in under a second. The 4B Base needs far more sampling steps and is meant for training, research, and cases where output variety matters more than latency.

Can I use FLUX.2 [klein] 4B Base commercially?

The weights are released under Apache 2.0, which permits commercial use of the model and its outputs. Check the license text and the maker's usage policy for any conditions that apply to your deployment.

Does it support image editing?

Yes. Black Forest Labs describes the klein family as a unified architecture for text-to-image and multi-reference editing, and the 4B Base inherits both abilities, so a single checkpoint can serve generation and edit requests alike.

Which GPU does it need for self-hosting?

The model card gives about 13GB of VRAM as the minimum, which covers cards in the RTX 3090 and 4070 class. That requirement only matters if you run the weights yourself.

How much does FLUX.2 [klein] 4B Base cost?

On TokenLab, FLUX.2 [klein] 4B Base costs $0.0019 per request. The pricing table above shows the full breakdown. Rates depend on the billing unit, specification, and usage. Compare matching conditions in the model's detailed pricing; a single rate does not determine the total cost.

Which endpoint should FLUX.2 [klein] 4B Base use?

Use https://api.tokenlab.sh/v1/images/edits for FLUX.2 [klein] 4B Base. The request example below shows the matching code shape.

Which operations does FLUX.2 [klein] 4B Base support?

FLUX.2 [klein] 4B Base supports Image editing, Text to image. Select an operation above to see its endpoint and request example.

Compare FLUX.2 [klein] 4B Base

Guides that use FLUX.2 [klein] 4B Base

Sources

Reviewed Oct 2, 2026

More from FLUX

Related models