Flux Dev
Black Forest Labs' development model. Fast, high-quality image generation with LoRA support.
Flux Dev is image generation AI model from Black Forest Labs, priced at €0.000 per 1M input tokens with a unknown context window.
Examples
See what Flux Dev can generate
Sample output
Landscape
Prompt: "A serene mountain lake at golden hour, with snow-capped peaks reflected perfectly in the still water, pine forests lining the shore, warm amber light casting long shadows across the meadow"
Sample output
Character Design
Prompt: "A cyberpunk street samurai with neon-blue cybernetic arm, wearing a weathered leather jacket, standing in a rain-soaked Tokyo alley with holographic signs reflecting in puddles"
Sample output
Still Life
Prompt: "An elegant flat lay arrangement of artisan coffee beans, a brass hand grinder, a ceramic pour-over dripper, and fresh croissants on a dark walnut table, morning light streaming from the left"
Pricing
API Integration
Use our OpenAI-compatible API to integrate Flux Dev into your application.
npm install railwailimport railwail from "railwail";
const rw = railwail("YOUR_API_KEY");
const images = await rw.run("flux-dev", "A beautiful sunset over Tokyo");
console.log(images[0].url);
// Or use the image() method for full control
const res = await rw.image("flux-dev", "A cat in space", {
size: "1024x1024",
n: 1,
});
console.log(res.data[0].url);Deep dive — Black Forest Labs's Flux Dev
Black Forest Labs was founded in August 2024 by Robin Rombach, Patrick Esser, Andreas Blattmann and Dominik Lorenz, the original Latent Diffusion and Stable Diffusion team from LMU Munich and Stability AI. Headquartered in Freiburg, Germany, the lab raised a $31M seed round led by Andreessen Horowitz and launched the FLUX.1 family in August 2024 with three tiers: FLUX.1 [pro] (closed, API-only, highest quality), FLUX.1 [dev] (open weights, non-commercial, near-pro quality) and FLUX.1 [schnell] (open weights Apache 2.0, distilled few-step).
Visit Black Forest Labs →FLUX.1 [dev] is the open-weights research variant of the FLUX.1 family, providing near-FLUX.1 [pro] quality with publicly downloadable weights. The architecture is a rectified-flow Diffusion Transformer (DiT) with approximately 12B parameters, trained with the flow-matching objective from the SD3 paper (Esser et al. 2024). It uses T5-XXL plus CLIP-L as parallel text encoders, allowing both long descriptive prompts and short keyword-style prompts. Image tokens come from a learned 16x latent VAE. Sampling typically uses 28-50 flow-matching steps with the Euler sampler at guidance ~3.5. FLUX.1 [dev] is guidance-distilled from a teacher and uses 'guidance' as a conditioning input rather than classifier-free guidance at inference, simplifying multi-step deployment. The model produces images up to 2 MP with strong prompt adherence, photorealistic detail and notably good in-image text rendering.
- Parameters
- ~12B parameters
- Context
- 512 tokens
- Near-state-of-the-art photorealism for an open model
- Up to 2 MP output resolution (1MP standard)
- Strong prompt adherence via T5-XXL
- Good in-image text rendering (short strings)
- Open weights — self-hostable, fine-tunable
- ControlNet, LoRA and IP-Adapter ecosystem
- Compatible with FLUX Tools (Fill, Canny, Depth, Redux)
- Best for: self-hosted creative pipelines, custom fine-tunes, research, ControlNet workflows.
Pretrained on a large licensed and publicly available image-text dataset; then distilled from a teacher to produce the [dev] guidance-distilled checkpoint. Exact composition and token counts are not disclosed.
License: FLUX.1 [dev] Non-Commercial License — free for research and personal use; commercial use requires a separate agreement with Black Forest Labs.
Known limitations
- Non-commercial license restricts production use
- Lower quality than FLUX.1 [pro] on fine details
- Slow on consumer GPUs without quantisation
- Hands and complex small text still imperfect
Frequently asked questions
Related Models
View all Image GenerationFlux 1.1 Pro Ultra
FLUX 1.1 Pro in ultra mode. Up to 4 megapixel images with raw mode for photorealism.
Google Imagen 4
Google's Imagen 4. Text-to-image with strong photorealism and improved typography support.
Google Imagen 4 Ultra
Premium Imagen 4 tier. Highest fidelity, prompt adherence and typography quality from Google.
Ideogram 3.0
Ideogram's flagship text-to-image model with industry-leading text rendering and prompt adherence.
Start using Flux Dev today
Get started with free credits. No credit card required. Access Flux Dev and 100+ other models through a single API.