Black Forest Labs

FLUX 1 Dev

Name: FLUX 1 Dev
Availability: InStock
Author: Black Forest Labs

FLUX 1 Dev is an open-weight text-to-image generation model from Black Forest Labs, positioned in the FLUX.1 family between the faster [schnell] variant and the hosted [pro] tier.

From

$0.025 / image

across 1 provider

Compare Prices

API Pricing

Provider	Price /img	Updated
Replicate	$0.025/img	8/18/2026

Prices updated daily. Last check: Aug 18, 2026

Model Details

General

Creator: Black Forest Labs
Modalities: Text

Capabilities

Tool Calling: No
Open Source: No
Aliases: black-forest-labs/flux-dev, FLUX.1 [dev], FLUX.1 [dev] FP8

Strengths & Limitations

Strengths

Open weights, so the model can be self-hosted, quantized, or fine-tuned rather than only accessed through a single vendor API
Available from multiple inference providers as well as on your own GPUs, giving pricing and deployment flexibility
Large LoRA and fine-tune ecosystem built on FLUX.1 [dev] for style, character, and product-specific adaptation
Strong reputation for prompt adherence and in-image text rendering compared to earlier open diffusion baselines
FP8 variant offered by some providers reduces VRAM requirements and shortens generation time
Standard image-generation controls (resolution, aspect ratio, guidance, step count, seed) support reproducible outputs
Sits mid-family: higher quality target than FLUX.1 [schnell] without requiring the hosted [pro] endpoints

Limitations

The FLUX.1 [dev] weights are released under a non-commercial license, so commercial deployment requires separate licensing or a provider with commercial terms
Slower per image than FLUX.1 [schnell], which is tuned for very low step counts
Text-to-image only — it is not a chat or reasoning model and does not produce text answers
Self-hosting a 12B-class image model needs a capable GPU; the FP8 build eases but does not remove this
Output quality is sensitive to guidance and step-count settings, so provider defaults can produce noticeably different results for the same prompt

Key Features

•Text-to-image generation from natural-language prompts

•Open, downloadable weights for self-hosting

•FP8 quantized variant for lower-memory inference

•Configurable inference step count and guidance scale

•Selectable output resolution and aspect ratio

•Seed control for reproducible generations

•LoRA and fine-tune compatibility for custom styles and subjects

•Availability through multiple hosted inference APIs under the black-forest-labs/flux-dev alias

About FLUX 1 Dev

FLUX 1 Dev (published as FLUX.1 [dev]) is a text-to-image diffusion model released by Black Forest Labs as part of the FLUX.1 family. Within that family it sits between FLUX.1 [schnell], which targets very low step counts and fast turnaround, and the hosted FLUX.1 [pro] endpoints. Because the weights are published, FLUX 1 Dev is offered both by inference APIs and by self-hosters, and it appears in our catalog under aliases including black-forest-labs/flux-dev and a quantized FLUX.1 [dev] FP8 variant. As an image generation model, FLUX 1 Dev is evaluated on prompt adherence, text rendering inside images, anatomy and composition quality, and generation latency at a given step count and resolution rather than on context window or token throughput. It accepts a text prompt and typically exposes controls such as output resolution and aspect ratio, guidance scale, inference step count, and a random seed for reproducibility. The FP8 build trades some numerical precision for reduced VRAM use and faster sampling, which is why some providers list it separately. In practice FLUX 1 Dev is a common default for general-purpose image generation work: concept art, marketing and social imagery, product mockups, and synthetic image data. It is also one of the more widely fine-tuned and LoRA-adapted base models in the open-weight image ecosystem, which makes it attractive when a team wants a specific visual style or subject baked in. Note that the FLUX.1 [dev] release carries a non-commercial license for the weights themselves; commercial use generally routes through Black Forest Labs' licensed offerings or a provider that has arranged commercial terms, so check the license and your provider's terms before shipping.

Common Use Cases

FLUX 1 Dev suits general-purpose image generation where quality matters more than the sub-second turnaround targeted by FLUX.1 [schnell]: concept art and storyboards, editorial and social imagery, product and packaging mockups, game and UI asset exploration, and synthetic training images for vision pipelines. Its in-image text handling makes it a reasonable choice for mockups that include signage, labels, or short headlines. Because the weights are open and widely adapted, it is a frequent base for LoRA-trained brand styles, recurring characters, or catalog-specific product looks — teams that need a consistent house style often prefer it over closed image APIs for that reason. For commercial pipelines, confirm licensing terms with Black Forest Labs or your chosen provider first; for the fastest possible drafts, pair it with a lower-step variant and reserve FLUX 1 Dev for final renders.

Frequently Asked Questions

How much does FLUX 1 Dev cost to use?

It depends on the provider. Image models are usually billed per generated image, sometimes scaled by resolution or inference steps, and some providers instead bill GPU time for a dedicated endpoint. Self-hosting shifts the cost to the GPU hourly rate. Because rates and billing models differ between providers, check the pricing table on this page for current figures.

What is FLUX 1 Dev best used for?

Text-to-image generation for concept art, marketing and social visuals, product mockups, game assets, and synthetic image data — plus as a base model for LoRA fine-tunes when you need a specific style, character, or product to render consistently.

How does FLUX 1 Dev differ from FLUX.1 [schnell] and FLUX.1 [pro]?

Within the FLUX.1 family, [schnell] is tuned for very few inference steps and the fastest generations, [dev] targets higher output quality at more steps, and [pro] is offered through Black Forest Labs' hosted endpoints. FLUX 1 Dev is the open-weight option in the middle of that range.

Can I use FLUX 1 Dev commercially?

The FLUX.1 [dev] weights are released under a non-commercial license. Commercial use generally requires a license from Black Forest Labs or use of a provider that has arranged commercial terms. Review the license and your provider's terms of service before deploying it in a paid product.

What is the FP8 variant?

FLUX.1 [dev] FP8 is a quantized build of the same model that stores weights at lower precision. It reduces VRAM requirements and typically speeds up generation, with some loss of numerical precision relative to the full-precision weights. Some providers list it as a separate, cheaper endpoint.

Does FLUX 1 Dev accept an input image?

Our metadata for this entry only confirms text-to-image generation. Image-to-image, inpainting, and control-style conditioning are commonly built on FLUX models by third parties, so check the specific provider's API reference for which endpoints and input types they expose.