FLUX 3 Klein is the small, fast member of Black Forest Labs' FLUX 3 family, with native 4K and up to 10 reference images. Paste a prompt list, run every line through Klein, and take the whole set back as one ZIP.
One per line — each line is a separate prompt.
Drag & drop or click to upload — PNG / JPEG / WEBP / GIF, up to 10MB each.
Black Forest Labs released FLUX 3 Image Klein on 15 October 2026. Everything below comes from BFL's own announcement and documentation, not from us; where a claim rests on BFL's internal testing, we say so.
BFL describes Klein as a compact, open-weights version of FLUX 3 Image with the same feature set: text-to-image, editing and multi-reference generation. It has 5.6 billion parameters, not counting the text encoder, and the version served over the API is the distilled checkpoint, built for low latency. A full-size FLUX 3 Image model sits above it for the hardest work; Klein is the one sized for volume, which is exactly what a batch is.
Klein replaces FLUX.2 [klein]. Against the 9B FLUX.2 Klein, BFL reports 38% fewer parameters and a gain of 91 Elo in text-to-image and 70 Elo in editing. In BFL's own blind evaluations it also ranks first against Qwen-Image 2.1, Krea 2, Ideogram V4.5 and FLUX.2 [pro]. Those are the lab's numbers, not an independent benchmark, so treat them as a reason to try FLUX 3 Klein on your own prompts rather than as a verdict.
Klein renders up to 4K natively and takes up to 10 reference images in one request, so a product, a character or a style board can travel through a whole batch. It also supports editing an existing image with a written instruction. The tiers, ratios and reference slots that apply here are in the table below, read from our live configuration rather than typed into this page.
Both Black Forest Labs models on BulkImagen, side by side. Ratios, resolutions, reference limits and credit prices come from the running model configuration, so the table always matches what the form accepts.
Everything in this table, the price included, is read from the live configuration, not from this page's copy.
The same three steps as every model on BulkImagen. What is specific to Klein sits in the middle step: how many references to attach, and which resolution tier to pay for.
Paste one prompt per line, or import a CSV. Only have an idea? Write it as one sentence and press Write with AI under the prompt box to turn it into one prompt per image. Plain descriptions work; Klein rewrites each line into a fuller prompt before it renders.
Attach up to the reference limit shown in the table if every image must keep the same product, face or style, then choose the aspect ratio and the 1K, 2K or 4K tier. The image count and total credits appear before anything runs.
Every prompt goes to FLUX 3 Klein in parallel. Retry any image that misses, then download the whole batch as one ZIP, each file named after its prompt.
Where a small, fast model with 4K output and a deep reference stack pays off when you generate in bulk.
Concept rounds, mood exploration, stock-style scenes, blog and newsletter images. When the job is a hundred options and a handful of keepers, the per-image price matters more than anything else, and FLUX 3 Klein is the lower-cost FLUX model on BulkImagen at every resolution tier. Compare the prices in the table above.
A product shot across a full catalog of backgrounds, or the same character through a storyboard. With up to ten reference images, Klein can hold the subject steady while each prompt changes the setting, the light or the pose, which is where single-image generators drift.
Posters, wall art, print-on-demand designs and hero banners that must stay sharp when enlarged. Klein renders 4K natively, so a whole series can come out at print resolution in one batch instead of being upscaled image by image afterwards.
Anything else? Email support@bulkimagen.com
Black Forest Labs' compact FLUX 3 model, a whole prompt list at a time: paste the lines, check the credits, take the ZIP.
