Generate Stunning Images in Under 1 Second

FLUX.2 Klein delivers state-of-the-art image quality at unprecedented speed. The 9B model matches outputs from models 5x its size, while the 4B runs on consumer GPUs with full commercial license.

By The Numbers

FLUX.2 Klein performance at a glance

Industry-leading speed and efficiency backed by real benchmarks

+1200

Faster Than Competitors

22 Million

Monthly Downloads

+500

Inference Steps

Seamless Integration

Works with your existing workflow

FLUX.2 Klein integrates with popular AI frameworks and platforms for effortless deployment.

Hugging Face Diffusers

Native support in the Diffusers library with optimized pipelines for both 4B and 9B models.

ComfyUI

Full ComfyUI integration for node-based workflows with custom sampling options.

Black Forest Labs API

Cloud API access for production deployments without managing infrastructure.

NVIDIA TensorRT

Optimized inference with TensorRT for maximum performance on NVIDIA GPUs.

Automatic1111 WebUI

Community extensions available for the popular Stable Diffusion WebUI.

Custom Pipelines

Build custom inference pipelines with full access to model weights and architecture.

Speed meets quality in one powerful model

The first image model to deliver production-quality results in under one second

Why FLUX.2 Klein

The first image model to deliver production-quality results in under one second

Generate 1024x1024 images in as fast as 0.5 seconds on GB200 hardware, or ~2 seconds on RTX 5090. 30% faster than any competing model while maintaining exceptional quality.

Model Variants

Choose the right model for your needs

FLUX.2 Klein comes in two sizes optimized for different use cases

Model Variants

FLUX.2 Klein comes in two sizes optimized for different use cases

• 4B Model - Apache 2.0 License
• 9B Model - Maximum Quality
• FP8 Quantized Versions
• Base Models for Fine-tuning

Technical Excellence

Built on cutting-edge architecture

Powered by Black Forest Labs’ latest rectified flow transformer technology

Rectified Flow Transformer

Advanced 9B parameter architecture that matches models 5x its size in output quality

4-Step Inference

Step-distilled to just 4 inference steps without sacrificing image fidelity

Qwen3 Text Encoder

8B parameter text embedder for superior prompt understanding and adherence

Multi-Reference Editing

Edit images using multiple reference inputs in a single unified workflow

C2PA Watermarking

Built-in content provenance with cryptographically signed metadata

Safety-First Design

Pre-training filtering and post-training fine-tuning against abuse vectors

Deploy Anywhere

From consumer GPUs to cloud infrastructure, FLUX.2 Klein adapts to your deployment needs.

Start generating for free

Choose the plan that fits your creative workflow

---

• 50 images per day

• 4B model access

• 1024x1024 resolution

• Community support

• 9B model access

• API access

• Commercial license

Popular

---

• 1,000 images per month

• 4B and 9B model access

• Up to 2048x2048 resolution

• Priority generation queue

• API access included

• Custom fine-tuning

• Dedicated support

---

• Unlimited generations
• All model variants
• Maximum resolution support
• Dedicated API endpoints
• Custom fine-tuning available
• Priority support
• SLA guarantee

FAQ

Frequently Asked Questions

Start Generating in Seconds

Join thousands of creators using FLUX.2 Klein for lightning-fast AI image generation