FLUX.2-Klein-4B
Black Forest Labs' compact, Apache 2.0 FLUX model, self-hosted by AutorunX as AX-KLN, delivers sub-second exact-identity generation at the platform's lowest cost.
What FLUX.2 Klein does
FLUX.2 Klein generates images from a text prompt and, when given multiple reference images, produces new compositions that preserve the exact identity of the subject in those references — a face, a product, a brand mark — across every generation.
On AutorunX it's the default model for Banner Ad creative and for exact-identity persona generation, the two jobs where its combination of low cost, fast turnaround, and multi-reference identity locking matters most.
Capabilities
Key features
- Multi-reference exact identity
- Takes multiple reference images of the same subject and generates new compositions that keep that subject's identity exact, which underpins AutorunX's persona-generation default.
- Sub-second generation speed
- A single compact 4B checkpoint is built for minimal latency, returning results far faster than larger, full-scale image models.
- Apache 2.0 open weights
- Fully open-weight with no commercial-use restrictions, which is what let AutorunX self-host it as AX-KLN rather than pay a per-call API fee.
- Low-VRAM self-hosting
- Runs on AutorunX's own 16GB VRAM lane, keeping the owned infrastructure footprint small relative to larger image models.
- Platform's lowest cost tier
- At 5 credits per image, it's tied with Z-Image Turbo as the cheapest image option on AutorunX, and roughly 10x cheaper per image than GPT Image 2.
How FLUX.2 Klein works
FLUX.2 Klein 4B is the compact entry point of Black Forest Labs' FLUX.2 model family, purpose-built for minimal latency rather than maximum scale. A single 4-billion-parameter checkpoint handles text-to-image generation, image editing, and multi-reference generation, instead of shipping separate models for each mode.
Its small footprint is the point: Klein 4B fits comfortably within a modest VRAM budget, which is exactly why AutorunX runs it in-house — as AX-KLN — on its own GPU fleet's 16GB VRAM lane instead of renting it from a third-party API. That ownership is what lets AutorunX offer it at close to the platform's lowest possible credit cost.
Multi-reference generation lets Klein take several input images (for instance, angles of the same person or product) and produce new outputs that stay exact-identity consistent with them, which is why AutorunX defaults to it for Face / Identity persona work where a generated character has to look like the same person across every shot.
What people use FLUX.2 Klein for
Default banner ad generation
Generate banner ad creative quickly and cheaply as the panel's default model, reserving pricier models for jobs that specifically need them.
Exact-identity persona shots
Produce Face / Identity persona images that hold the same face and identity consistent across a full photoshoot's worth of generations.
High-volume campaign creative
Generate large batches of on-brand creative at low per-image cost for campaigns that need many variations.
Budget-conscious prototyping
Iterate on creative direction cheaply before committing credits to a higher-cost model for final hero assets.
How to use FLUX.2 Klein on AutorunX
FLUX.2 Klein is the default model for Banner Ad, in the Image module.
- 01
Open Banner Ad in Image
FLUX.2 Klein is the default model here — head to the Banner Ad service.
- 02
Set up your input
Write your prompt, and for exact-identity persona work, upload multiple reference images of the same subject.
- 03
Confirm FLUX.2 Klein in the model picker
It's pre-selected by default; open the picker if you want to compare it against another model.
- 04
Generate
Run the generation — expect a fast turnaround given the model's compact, low-latency design.
- Billed per image from your shared AutorunX credit wallet.
- 5 credits / image — the cheapest image option on AutorunX alongside Z-Image Turbo, and roughly 10x cheaper than GPT Image 2's per-image cost
Tips for better results
- Feed it multiple angles for identity locks
- Provide several reference images of the same subject rather than one — Klein's multi-reference mode uses the extra views to hold identity steadier across outputs.
- Trust it as the default, not a fallback
- It's AutorunX's default for Banner Ad and persona generation for a reason — reach for a pricier model only when a specific shot needs capabilities Klein doesn't cover.
- Keep references consistent in lighting
- Reference images shot under similar lighting produce more coherent identity-locked outputs than mismatched sources.
- Use it for volume, escalate for hero shots
- Its low cost makes it ideal for generating many campaign variations; move only the winning direction to a higher-fidelity model if needed.
FLUX.2 Klein — frequently asked
Related models
Z-Image Turbo
Alibaba Tongyi LabAlibaba Tongyi Lab's 6B open-weight model generates photorealistic images in as few as 8 steps, making it AutorunX's cheapest image option.
Qwen-Image-Edit
AutorunX ownedAlibaba's open-weight 20B image editor, self-hosted by AutorunX as AX-QWN, preserves identity through fine-grained edits at near-zero marginal cost.
Nano Banana
Google DeepMindGoogle's Gemini-family image model, nicknamed Nano Banana for its viral photorealistic edits, generates and edits images up to a 4K-capable tier.
Ready to create with FLUX.2 Klein?
Open Banner Ad and generate your first image in a couple of minutes.
Free plan, no card required.