Black Forest Labs offers the FLUX AI image generation models through a usage-based API and Playground, with open-weights licences for teams that host their own.
Black Forest Labs generates finished pictures from a text prompt or a reference photo, and lets developers call that from their own software. Its FLUX models run on the company's servers, so no one needs GPUs of their own to use them. FLUX 3 Image is the current flagship, and the FLUX.2 and FLUX.1 lines remain on offer beside it. The company, founded in 2024, builds the models and runs the service itself. FLUX 3 Image handles text-to-image and image-to-image work, editing specific details while blending several reference images into one result. The FLUX.2 family runs from the compact [klein] versions to [pro], [flex] and [max], and it keeps a character's face, clothing and style steady across a series. The older FLUX.1 line adds Kontext and FLUX.1 Fill, which repaints a masked part of an image. Among the FLUX Image Tools, Outpainting extends a picture onto a larger canvas or a social format. FLUX VTO handles virtual try-on. Developers and product teams who want image generation inside their own apps are the main users, along with creators working in the browser Playground. Access is bought as prepaid credits with no subscription or seat fee, and each image is charged by model and output resolution. Companies that would rather host the models license open weights, sized by monthly image volume, domains and users. Larger buyers agree volume pricing, service levels and combined API and weights deals with sales. Product photography, mockups and posters with legible typography are typical FLUX jobs, as is resizing one campaign visual into every social format. Developers working in Claude, Cursor or Codex reach the models through the FLUX MCP server, which returns several variations of a prompt at once. Enterprises already on Microsoft's cloud can use the models through Azure AI Foundry. Machine-learning teams train LoRAs on the [klein] weights to teach the model a product or a house style. A synthetic data licence lets AI teams use the generated images as training material for models of their own.
Pricing
View plans-
Pay as you go: FLUX 3 Image
Custom pricing
- $0.041 per image at 768 by 768, usage-based from prepaid credits at $0.01 each; no monthly or annual billing, no seat fee
- List price per image (USD): $0.041 at 768 by 768, $0.048 at 1K, $0.07 at 1.5K, $0.10 at 2K, $0.607 at 4K
- Same price in the API and the Playground
- Reference images and prompts cost nothing extra; rejected, failed and moderated requests are not billed
- Reference images up to 16 MP; anything above 4 MP is downscaled to 4 MP
- No negative prompt field and no seed parameter on the FLUX 3 Image endpoint
-
Pay as you go: FLUX.2 and FLUX.1 models
Custom pricing
- From $0.014 per image (FLUX.2 [klein] 4B), usage-based from prepaid credits; price per image at entry resolution, rising with output megapixels
- Per image (USD): FLUX.2 [klein] 4B $0.014, [klein] 9B $0.015, [pro] $0.03, [flex] $0.05, [max] $0.07
- Per image (USD): FLUX.1 Kontext [pro] $0.04, Kontext [max] $0.08
- A batch costs the base price times the number of images, so 4 FLUX.2 [pro] images from $0.12
- Seed parameter for reproducible results on FLUX.2 endpoints
- Optional auto top-up of the credit balance
-
Open Weights: Builder
Custom pricing
- Self-hosting licence bought online; price not listed
- FLUX.2 [klein] models with fine-tuning and LoRA rights
- Up to 10K images a month, 1 domain, 10 licensed users
-
Open Weights: Platform
Custom pricing
- Self-hosting licence bought online; price not listed
- FLUX.2 [klein] 9B and FLUX.2 [dev] with fine-tuning and LoRA rights
- Up to 100K images a month, 1 domain, 10 licensed users
Features
- Included: Text-to-image generation
- Included: Image-to-image transformation
- Included: Inpainting and generative fill
- Included: Outpainting canvas extension
- Not included: Image upscaling and enhancement
- Not included: Background removal
- Not included: Negative prompt control
- Included: Seed and reproducibility control
- Included: Style presets and reference images
- Included: Custom model or LoRA training
- Included: Batch image generation
- Included: Commercial use rights on paid plans
- Included: Content credentials or watermarking
- Included: Aspect ratio and resolution presets
- Included: Image generation API
- Not included: Prompt history and asset gallery
- Included: Face and character consistency
- Included: Adult content filtering
Additional Features
- FLUX 3 Image text-to-image and image-to-image
- Editing with several combined reference images
- FLUX.2 [klein], [pro], [flex] and [max] models
- FLUX.1 Kontext models
- FLUX.1 Fill masked inpainting
- FLUX Outpainting with social format presets
- FLUX VTO virtual try-on
- Character identity held across a series of images
Best for
- Developers adding image generation to an app through an API
- Product teams that want per-image billing instead of a subscription
- Companies running image models on their own servers under licence
- E-commerce teams producing product shots and mockups
- Designers who steer output with reference images
- Marketing teams resizing one visual into many social formats
- Machine-learning teams fine-tuning a model on a product or style
- Developers who work in MCP clients such as Claude, Cursor or Codex
Use cases
- Generating product photography and mockups
- Designing posters with legible typography
- Extending an image to fit a social media format
- Editing details of an existing photo from a prompt
- Combining several reference images into one scene
- Keeping one character consistent across a campaign
- Repainting a masked region of an image
- Virtual try-on of clothing
- Training a LoRA on a brand's product or style
- Adding image generation to a SaaS product
- Producing synthetic training data for AI models
- Generating image variations from an AI coding assistant
Screenshots & Videos
Explore Black Forest Labs in action
User Reviews
Compare with Similar Software
Click to compare side by side with Black Forest Labs
