FLUX AI models: image generation & editing via API

The FLUX family from Black Forest Labs covers everything from fast prototyping to premium image generation and editing. Run any of these models with one line of code.

FLUX 2 — Latest generation

FLUX.2 Max

The highest-fidelity model in the FLUX family. Use it for:

Product photography and e-commerce imagery
Character consistency across batches (up to 8 reference images)
Precise color control with hex codes
Cinematic key frames and 3D view generation

Try FLUX.2 Max →

FLUX.2 Pro

High-quality generation and editing at a lower price than Max. Good for:

Structured JSON prompting for precise control over camera, lighting, and composition
Multi-reference editing with up to 8 images
High-volume production work
Text rendering and infographics

Try FLUX.2 Pro →

FLUX.2 Flex

The typography and design specialist. Good for:

Clean text, captions, and complex layouts
Memes, posters, infographics, and UI mockups
Adjustable quality-speed tradeoff (1-50 steps)
Up to 10 reference images for multi-image blending

Try FLUX.2 Flex →

FLUX.2 Klein 4B

Sub-second generation for production and real-time apps. Good for:

Near real-time applications and live previews
Consistent characters and products across batches (up to 5 reference images)
Text rendering and spatial reasoning
Running on consumer hardware

Try FLUX.2 Klein 4B →

Image editing

FLUX Kontext Pro & Max

Text-based image editing — describe what you want changed and it happens. Good for:

Style transfer (photo → watercolor, oil painting, sketch)
Changing clothing, hair, backgrounds, and objects
Replacing text in signs and labels
Character consistency across edits

Try Kontext Pro → · Try Kontext Max →

FLUX.1 Fill, Depth & Canny

Structure-preserving editing tools:

FLUX.1 Fill — inpainting and outpainting, removing or replacing objects
FLUX.1 Depth — retexturing while keeping 3D relationships
FLUX.1 Canny — edge-preserving style transfer and sketch-to-image

Previous generation

These older models are still useful for specific workflows:

FLUX.1 Schnell — the fastest FLUX model, great for real-time apps and quick iteration
FLUX.1 Dev — open-source 12B parameter model for learning, prototyping, and self-hosting
FLUX1.1 Pro Ultra — up to 4 megapixel images with "raw" mode for natural realism

Make it your own

Fine-tune FLUX models with your own images to create consistent characters, match your brand style, or build specialized generators.

Try it out

Compare FLUX models side by side in the playground.

Open the playground →

Questions? Join us on Discord.

Frequently asked questions

Which FLUX model should I start with?

For most people, black-forest-labs/flux-2-pro is the best starting point. It delivers high-quality images with strong prompt adherence, supports up to 8 reference images, and handles both generation and editing. If you need the absolute highest fidelity, step up to black-forest-labs/flux-2-max.

Which models are the fastest?

black-forest-labs/flux-2-klein-4b generates images in sub-second time — it's a 4 billion parameter model distilled from the larger FLUX.2 base. Great for production apps and real-time use cases. black-forest-labs/flux-schnell is another fast option from the previous generation.

What works best for editing existing images?

black-forest-labs/flux-kontext-pro and black-forest-labs/flux-kontext-max let you edit images with natural language — change clothing, swap backgrounds, replace text, or transfer styles. For structure-preserving edits, use black-forest-labs/flux-fill-pro (inpainting), black-forest-labs/flux-depth-pro (depth-guided), or black-forest-labs/flux-canny-pro (edge-guided).

What's the best model for typography and text in images?

black-forest-labs/flux-2-flex is the typography specialist — it reliably renders clean text, captions, and complex layouts. black-forest-labs/flux-kontext-max also handles text well, especially for editing text within existing images.

What's the difference between FLUX 2 and FLUX 1?

FLUX 2 models (Max, Pro, Flex, Klein 4B) are the current generation with better image quality, multi-reference support, structured JSON prompting, and hex color control. FLUX 1 models (Schnell, Dev, 1.1 Pro Ultra) are the previous generation — still useful for speed (Schnell), open-source work (Dev), and large images (Pro Ultra), but FLUX 2 is better for most new projects.

Can I fine-tune FLUX models?

Yes. Use black-forest-labs/flux-pro-trainer to train a LoRA on your own images. You can also fine-tune Kontext models with replicate/fast-flux-kontext-trainer. Browse existing fine-tunes in the FLUX fine-tunes and Kontext fine-tunes collections.

Can I use FLUX models for commercial work?

Yes — most FLUX models are licensed for commercial use. Always check the license on the model page, particularly for the open-weight Dev models.

How do I get consistent characters across multiple images?

FLUX 2 models support multi-reference inputs. black-forest-labs/flux-2-max accepts up to 8 reference images and maintains character consistency across batches. black-forest-labs/flux-2-klein-4b supports up to 5 references for faster generation.