FLUX (Black Forest Labs)
State-of-the-art text-to-image models from the creators of Stable Diffusion, offering the FLUX.1 family with Pro, Dev, and Schnell variants for photorealistic generation
Overview
FLUX is a family of state-of-the-art text-to-image models developed by Black Forest Labs, founded by the original creators of Stable Diffusion (Robin Rombach, Andreas Blattmann, and team). The FLUX.1 models represent a significant leap in image generation quality, excelling at photorealism, prompt adherence, and text rendering within images. The family includes three variants: FLUX.1 Pro (highest quality, API-only), FLUX.1 Dev (open-weights for non-commercial use), and FLUX.1 Schnell (fastest, Apache 2.0 licensed). Built on a hybrid architecture combining multimodal and parallel diffusion transformer blocks with flow matching, FLUX has quickly become the benchmark for image quality, consistently ranking #1 on community leaderboards like Artificial Analysis.
The Verdict
Who Should Use FLUX?
Best For
- Developers needing best-in-class image quality via API
- Creative professionals requiring photorealistic outputs
- Projects needing accurate text rendering in images
- Researchers and hobbyists (Dev/Schnell open weights)
- Applications requiring diverse human representation
- Teams wanting Stable Diffusion alternatives with better quality
Not Ideal For
- Users wanting a consumer-friendly UI (use Midjourney)
- Those needing video generation (image-only for now)
- Budget-constrained projects (Pro API costs add up)
- Commercial use without API (Dev is non-commercial only)
- Users uncomfortable with technical setup for local models
What's Great
- Industry-leading photorealism and prompt adherence
- Excellent text rendering within generated images
- Open-weights models available (Dev and Schnell)
- Founded by Stable Diffusion creators—proven expertise
- Fast inference with Schnell (1-4 steps)
- Strong human anatomy and diverse representation
- Multiple API providers (Replicate, fal.ai, Together)
- Active development with frequent improvements
Watch Out For
- No official consumer UI—requires API or third-party tools
- Pro model is API-only, not available for local use
- Dev model restricted to non-commercial use
- Higher VRAM requirements than SD 1.5 (12B parameters)
- Relatively new company (founded 2024)
- Pricing through various providers can be confusing
- Less community tooling than Stable Diffusion ecosystem
Pricing
View all features & details
Model Variants
- FLUX.1 Pro — Highest quality, API-only
- FLUX.1 Pro 1.1 — Improved version with better coherence
- FLUX.1 Pro Ultra — 4MP output (2048x2048+)
- FLUX.1 Dev — Open weights, non-commercial
- FLUX.1 Schnell — Fast (1-4 steps), Apache 2.0
- FLUX.1 Fill — Inpainting and outpainting
- FLUX.1 Canny/Depth — ControlNet variants
- FLUX.1 Redux — Image variation model
Technical Specs
- 12 billion parameters
- Hybrid architecture (MMDiT + parallel DiT)
- Flow matching training
- Rotary positional embeddings
- T5 + CLIP text encoders
- Native 1024x1024 resolution
- Up to 4MP with Ultra variant
- FP8/FP16/BF16 inference support
API Providers
- Black Forest Labs API (api.bfl.ml)
- Replicate
- fal.ai
- Together AI
- Fireworks AI
- ComfyUI (local)
- Automatic1111 (community)
Key Capabilities
- Photorealistic generation
- Accurate text rendering
- Diverse human representation
- Complex scene composition
- Strong prompt adherence
- Inpainting and outpainting (Fill)
- ControlNet support
- Image variations (Redux)
Company & Funding
Company Background
- Founded: August 2024
- Founders: Robin Rombach, Andreas Blattmann, et al.
- Previously created Stable Diffusion at Stability AI
- Headquarters: Freiburg, Germany
Funding & Growth
- $31M Series A (August 2024)
- Led by Andreessen Horowitz
- General Catalyst, Dell Technologies participated
- Valued at over $100M
How It Compares
| Feature | FLUX.1 Pro | Midjourney V6 | DALL-E 3 | Stable Diffusion XL |
|---|---|---|---|---|
| Image Quality | Best photorealism | Best aesthetics | Very good | Good |
| Text Rendering | Excellent | Good | Excellent | Limited |
| Prompt Adherence | Excellent | Good | Very good | Good |
| API Access | Yes | No | Yes | Yes |
| Open Weights | Dev/Schnell only | No | No | Yes |
| Local Hosting | Dev/Schnell | No | No | Yes |
| Speed | Fast (Schnell: 1-4 steps) | 30-60s | 10-20s | Variable |
| Price per Image | ~$0.05 | $0.05-0.20 | $0.04 | Free (local) |
| Best For | Quality + API | Artistic style | Ease of use | Customization |