FLUX.2 Dev
Black Forest Labs’ FLUX.2 Dev generates and edits images from text descriptions. This 32 billion parameter model combines image generation and editing in one, letting you create new images from scratch, modify existing ones, or blend multiple reference images together.
What it does
FLUX.2 Dev handles three main tasks:
Generate images from text
Describe what you want and the model creates it. The model follows complex prompts well, understands real-world lighting and spatial relationships, and can render readable text in images.
Edit existing images
Upload an image and describe what you want to change. The model can modify specific parts while keeping the rest intact, working with images up to 4 megapixels.
Combine multiple references
Use up to 10 reference images simultaneously. The model maintains character consistency, product details, or style elements across all your references without needing additional training.
When to use it
This model works well for:
- Product photography and mockups where you need consistent branding and lighting
- Character designs that stay consistent across multiple scenes
- Infographics and UI mockups with complex typography
- Concept art and visualization that needs photorealistic detail
- Marketing materials that blend multiple reference images
The dev variant is optimized for efficiency through guidance distillation, making it faster than typical models of this size while maintaining high quality output.
Technical details
FLUX.2 Dev uses a latent flow matching architecture that combines the Mistral-3 24 billion parameter vision-language model with a rectified flow transformer. The vision-language model brings real-world knowledge and contextual understanding, while the transformer handles spatial relationships, materials, and composition.
The model trains on a new variational autoencoder designed to balance quality, learnability, and compression. This lets the model generate sharper textures, more stable lighting, and better handle fine details compared to earlier approaches.
Improvements over FLUX.1
FLUX.2 Dev includes several upgrades:
- Multi-reference support for maintaining consistency across up to 10 input images
- Improved text rendering that reliably produces legible typography
- Better prompt following for complex, structured instructions
- Enhanced photorealism with sharper textures and more accurate lighting
- Higher resolution support up to 4 megapixels for editing
- Stronger grounding in real-world knowledge and spatial logic
License and usage
FLUX.2 Dev is available for commercial use on Replicate. The weights are open and available to drive new research and help artists develop innovative workflows. See the API documentation for details.
Try it
You can try FLUX.2 Dev on the Replicate playground.