Back to FLUX
Black Forest Labs /

FLUX.1 schnell
FLUX.1 schnell

Open WeightsImage GentextimageUpdated August 1, 2024
Some details or benchmark scores on this page are self-reported by developers and unconfirmed.

FLUX.1 schnell

Model Overview

FLUX.1 [schnell] is a highly optimized, 12-billion parameter text-to-image model developed by Black Forest Labs. Released under an open Apache 2.0 license, it leverages a rectified flow transformer architecture to deliver state-of-the-art visual generation capabilities. Unlike standard diffusion models that require dozens of steps, FLUX.1 [schnell] is specifically designed for high-speed inference, producing professional-quality outputs in just 1 to 4 steps.

Capabilities

  • Ultra-Fast Generation: Achieves high-quality image generation in only 1–4 steps, thanks to latent adversarial diffusion distillation.
  • High Visual Fidelity: Despite its rapid generation speed, the model maintains competitive visual quality and excellent text rendering capabilities.
  • Prompt Adherence: Exceptional ability to accurately follow complex text prompts, including spatial awareness and specific style requests.
  • Open Weights: Full access to the model weights for self-hosting, fine-tuning, and integration into custom applications.

Example Use Cases

  • Rapid Prototyping: Quickly iterating through visual concepts for design, advertising, and content creation.
  • Real-Time Applications: Integration into chatbots, interactive applications, or workflows that require near real-time image generation.
  • Cost-Efficient Batch Processing: Generating large volumes of images for datasets, synthetic data generation, or e-commerce at a very low cost per image.
  • Local Deployment: Running high-quality image generation locally on consumer GPUs for privacy-focused or offline tasks.

Performance & Benchmarks

  • Throughput: Capable of generating thousands of images per hour on modern hardware (e.g., RTX 4090 or H100). Benchmarks indicate it can generate approximately 5,243 images per dollar on cloud platforms.
  • Quality Score: While official standard benchmarks are not published by Black Forest Labs, community evaluations estimate its CLIP score around 0.32, matching or exceeding many closed-source alternatives.
  • Efficiency: The 1-4 step inference drastically reduces compute requirements compared to its counterparts like FLUX.1 [dev] and [pro].

Intended Use & Limitations

FLUX.1 [schnell] is intended for personal, scientific, and commercial use. It is ideal for developers and creators needing fast, cost-effective image generation. Limitations:

  • Being optimized for speed (1-4 steps), it may occasionally lack the extreme micro-details found in the heavier, slower FLUX.1 [pro] model.
  • As an open-weights model, users are responsible for implementing appropriate safety filters to prevent the generation of harmful or policy-violating content.

About Black Forest Labs

Black Forest Labs is an AI research organization dedicated to building advanced, accessible foundational models for generative media. They focus on providing open, high-quality models to the developer community, emphasizing both performance and permissive licensing.

Key Features

12B rectified flow transformer architecture

Feature 01

Apache 2.0 license for commercial use

Feature 02

Fast 4-step generation with high visual fidelity

Feature 03

Exceptional prompt adherence and text rendering

Feature 04

You might also want to compare

Verified Sources

Tags

imagediffusionopen-source

Model Specs

open-weights

Parameters

12B

Context Window

undisclosed

License

Apache 2.0

Deployment

self-hostable

Resources & Links

Lineage

Model Family

Part of the FLUX family

Only release in this line currently tracked.

Curator Notes

Standardized FID/CLIP benchmarks are not officially published by BFL; the score listed is an estimate from community evaluations.

Compare Specs

Compare parameters, context windows, modalities, and benchmark scores of this model side-by-side with others.

Compare Model