// AI NEWS · MODEL RELEASE

Black Forest Labs Open-Sources FLUX 3 Image Klein, a 5.6B Image Model That Runs on an 8GB GPU

Released October 15, 2026, FLUX 3 Image Klein is a 5.6B-parameter open-weights model for text-to-image, editing, and multi-reference generation — small enough to run locally on a consumer GPU, with native 4K output and bounding-box layout control carried over from the full FLUX 3 Image.

2026-10-16 · by Moe Ameen

What happened

Black Forest Labs released FLUX 3 Image Klein on October 15, 2026 — an open-weights image model that brings the capabilities of its FLUX 3 Image system down to a size that runs on a single consumer GPU. The lab describes Klein as a 5.6B-parameter model for text-to-image generation, image editing, and multi-reference generation (parameter counts cover the image generator, not the text encoders). It is the second open-weights release in the FLUX 3 family, after FLUX 3 Action.

The point of Klein is local, low-latency, self-hosted generation. It ships in two forms — a base model meant for fine-tuning and a step-distilled model built for fast inference — plus FP8, INT8, and NVFP4 builds developed with NVIDIA. The generator weights need about 11.2 GB in BF16 and 5.6 GB in FP8, and Black Forest Labs says the INT8 build runs on NVIDIA GeForce RTX cards with as little as 8 GB of VRAM through ComfyUI or Diffusers. On an RTX 5090 the distilled INT8 model renders a 1-megapixel image in roughly 4 seconds, and under 9 seconds on an RTX 4070. Weights and reference code are on Hugging Face and GitHub, and the distilled model is also on the BFL API as flux-3-image-klein, starting at $0.009 per 1,000 images; there is day-one support in ComfyUI, Hugging Face Diffusers, and several serving frameworks.

Klein inherits the core features of the full FLUX 3 Image: native 4K generation, pixel-perfect editing, multi-reference generation with up to 10 reference images, and bounding boxes to place elements or confine an edit to a defined region. Black Forest Labs reports that in its own evaluations Klein ranks first among open-weights models in text-to-image, multi-reference generation, and image editing, and that it beats the earlier FLUX.2 [klein] 9B model on internal Elo while using 38% fewer parameters. Those are the lab's own numbers — independent leaderboard results were not available at release.

The license is the FLUX Kommunity License. It permits free commercial use for individuals and for companies with annual revenue under $5 million, along with research and non-commercial use; larger organizations deploy commercially through self-hosted licensing or the API. Confirm current pricing, VRAM figures, and license terms against Black Forest Labs' own pages before relying on them, since early releases move.

Why it matters for creators

  • Open weights that fit an 8 GB consumer GPU change the economics: once the model runs on hardware you already own, the marginal cost of a render drops toward the cost of electricity, so generating at volume stops being a per-image API bill.
  • Local and on-device inference means the images — and any sensitive source references — never leave your machine, which matters for client work, unreleased products, and anyone with data-sovereignty constraints.
  • The base model is built for fine-tuning, and LoRA training is supported through Diffusers, Ostris’s AI Toolkit, and NVIDIA NeMo AutoModel — so a brand can train the model on its own look rather than fighting a prompt for consistency.
  • The FLUX Kommunity License clears free commercial use for individuals and sub-$5M companies, so most independent creators and small businesses can use Klein commercially without a separate license.
  • It is still a raw model, not an app: Klein outputs an image file with no caption, no per-platform crop, no surrounding copy, and no way to schedule or publish — the distribution work is all yours.

How to act on this with Kompozy

The news here is that high-quality image generation just became something you can run for free on your own GPU — but a folder of locally rendered 4K stills is raw material, not a content calendar. The practical move the day this lands is to pair Klein's near-zero marginal cost with an engine that turns renders into published posts. Fine-tune a LoRA on your brand's look, batch-generate on-brand stills locally at whatever volume you want, then feed them into Kompozy as sources. From one still, Kompozy renders a brand-exact multi-slide Carousel through HyperFrames, builds a Photo Post and a Quote Graphic from the same image, and writes platform-native captions in your voice through the Persona Brief.

That pairing is what converts a cost advantage into a distribution advantage. Klein makes the pixels cheap and private; Kompozy adds the formats a still model never produces — a Persona Short or Listicle Video spun from the same concept, plus a Blog Article and an Email Newsletter — and then schedules and publishes the whole set across eight social platforms plus blog and email from one queue with Autopilot and a per-post review pass. Generate locally and free, publish everywhere on brand: that is the workflow the open-weights release unlocks.

Quick takeaways

  • FLUX 3 Image Klein launched October 15, 2026 as a 5.6B-parameter open-weights image model from Black Forest Labs — the second open-weights release in the FLUX 3 family.
  • Its INT8 build runs on an 8 GB NVIDIA RTX GPU via ComfyUI or Diffusers; the distilled model renders a 1 MP image in about 4 seconds on an RTX 5090.
  • It carries over the full FLUX 3 Image features: native 4K, pixel-perfect editing, up to 10 reference images, and bounding-box layout control.
  • Weights are on Hugging Face (base + distilled); the FLUX Kommunity License allows free commercial use for individuals and companies under $5M in annual revenue.

Frequently asked questions

What is FLUX 3 Image Klein?

FLUX 3 Image Klein is Black Forest Labs’ open-weights image model, released October 15, 2026. It is a 5.6B-parameter model for text-to-image generation, image editing, and multi-reference generation, small enough to run locally on a consumer GPU while keeping the full FLUX 3 Image feature set: native 4K output, pixel-perfect editing, up to 10 reference images, and bounding-box layout control.

What hardware does FLUX 3 Image Klein need?

Black Forest Labs says the INT8 build runs on NVIDIA GeForce RTX GPUs with as little as 8 GB of VRAM through ComfyUI or Diffusers. The generator weights are about 11.2 GB in BF16 and 5.6 GB in FP8. On an RTX 5090 the distilled INT8 model renders a 1-megapixel image in roughly 4 seconds, and under 9 seconds on an RTX 4070.

Is FLUX 3 Image Klein free to use commercially?

Under the FLUX Kommunity License, Klein is free for commercial use by individuals and by companies with annual revenue under $5 million, as well as for research and non-commercial use. Larger organizations deploy commercially through self-hosted licensing or the BFL API, where the distilled model starts at $0.009 per 1,000 images. Confirm the current license terms on Black Forest Labs’ pages.

How is FLUX 3 Image Klein different from FLUX 3 Image?

FLUX 3 Image is the full-size API model; Klein is the open-weights, 5.6B version built to run locally and at low latency. Klein shares FLUX 3 Image’s core capabilities — native 4K, pixel-perfect editing, up to 10 reference images, and bounding boxes — but trades the largest model’s scale for self-hosting, fine-tuning, and on-device speed on consumer hardware.

Related news

← All AI news · Get started →