B

FLUX.1 [schnell] - Black Forest Labs

Black Forest Labs
Vision
N/A
Context
N/A
Input /1M
N/A
Output /1M
N/A
Max Output
Run API Call Open in Chat

FLUX.1 [schnell] - Black Forest Labs

Model Overview

Property Value
Model Name FLUX.1 [schnell]
Model ID black-forest-labs/FLUX.1-schnell
Creator Black Forest Labs
Type Text-to-Image Diffusion Model
Architecture Rectified Flow Transformer
Parameters 12 billion
License Apache 2.0
Training Method Latent Adversarial Diffusion Distillation

Description

FLUX.1 [schnell] is Black Forest Labs' fastest image generation model, a 12 billion parameter rectified flow transformer capable of generating high-quality images from text descriptions. Trained using latent adversarial diffusion distillation, it delivers cutting-edge output quality and competitive prompt following that matches the performance of closed-source alternatives.

The model is specifically tailored for local development and personal use, offering lightning-fast inference with sub-second generation times in just 1-4 denoising steps.

Pricing

Replicate

  • Per Image: ~$0.003 per image ($3 per 1,000 images)
  • Per Second Runtime: $0.001525/second
  • Median Cost (p50): $0.0015

fal.ai

  • Per Megapixel: $0.003
  • Billing: Rounded up to nearest megapixel
  • No subscription fees or minimum commitments

Supported Parameters

Core Parameters

Parameter Type Default Range/Options Description
prompt string required - Text description for image generation
num_inference_steps integer 4 1-4 Denoising steps (4 recommended, lower = faster but lower quality)
guidance_scale float 0.0-3.5 0.0-20.0 CFG scale for prompt adherence
seed integer random any Random seed for reproducible generation
num_outputs integer 1 1-4 Number of images to generate
max_sequence_length integer 256 - Maximum prompt token length

Image Size Parameters

Parameter Type Default Options
aspect_ratio string 1:1 1:1, 16:9, 21:9, 3:2, 2:3, 4:5, 5:4, 3:4, 4:3, 9:16, 9:21
megapixels string 1 0.25, 1
image_size string square square_hd, square, portrait_4_3, portrait_16_9, landscape_4_3, landscape_16_9

Output Parameters

Parameter Type Default Options Description
output_format string webp webp, jpg, png Image file format
output_quality integer 80 0-100 JPEG/WebP quality (PNG ignores this)

Performance Parameters

Parameter Type Default Description
go_fast boolean true Use fp8 quantized model for faster inference
acceleration string - Speed optimization: "none", "regular", "high"
enable_safety_checker boolean true Content safety filtering

Key Features

  • Ultra-Fast Inference: Generates high-quality images in only 1-4 steps with sub-second results
  • Cutting-Edge Quality: Output quality matches closed-source alternatives
  • Competitive Prompt Following: Strong adherence to text prompts
  • Flow-Based Distillation: Optimized for ultra-fast inference
  • Commercial Ready: Apache 2.0 license allows personal, scientific, and commercial use

Limitations

  • Not Factual: Not intended to provide factual information
  • Societal Biases: May amplify existing societal biases
  • Prompt Sensitivity: Prompt following heavily influenced by prompting style
  • Failure Cases: May fail to generate output matching prompts in some cases

API Endpoints

Replicate

  • Endpoint: https://api.langmart.ai/v1/predictions
  • Model: black-forest-labs/flux-schnell

fal.ai

  • Endpoint: fal-ai/flux/schnell
  • SDK: @fal-ai/client

BFL API (Official)

FLUX.1 Family

  • FLUX.1 [dev]: Higher quality, non-commercial license
  • FLUX.1 [pro]: Production-grade via BFL API
  • FLUX.1 Fill [dev]: In/out-painting capabilities
  • FLUX.1 Canny [dev]: Edge detection conditioning
  • FLUX.1 Depth [dev]: Depth map conditioning
  • FLUX.1 Redux [dev]: Image variation generation
  • FLUX.1 Kontext [dev]: In-context image editing

FLUX.2 Family (Latest)

  • FLUX.2 [max]: Highest capability, best editing consistency
  • FLUX.2 [pro]: Production-grade, 4MP resolution
  • FLUX.2 [flex]: Accessible variant
  • FLUX.2 [klein]: Apache 2.0 licensed

Performance Metrics

Metric Value
Inference Steps 1-4 (4 recommended)
Generation Time ~0.83 seconds (typical)
Total Run Time ~0.84 seconds
Hardware NVIDIA H100 GPU (Replicate)
Data Precision bfloat16 (bf16)
Total Runs 576.4M+ (Replicate)

Hardware Requirements

  • Recommended GPU: NVIDIA with CUDA support
  • VRAM: Can use CPU offloading to reduce VRAM requirements
  • Data Type: bfloat16 recommended for optimal performance
  • Local Deployment: Suitable for local development environments

Usage Examples

Python (Diffusers)

import torch
from diffusers import FluxPipeline

# Load the model
pipe = FluxPipeline.from_pretrained(
    "black-forest-labs/FLUX.1-schnell",
    torch_dtype=torch.bfloat16
)
pipe.enable_model_cpu_offload()  # Reduce VRAM usage

# Generate image
prompt = "A cat holding a sign that says hello world"
image = pipe(
    prompt,
    guidance_scale=0.0,
    num_inference_steps=4,
    max_sequence_length=256,
    generator=torch.Generator("cpu").manual_seed(0)
).images[0]

image.save("flux-schnell.png")

JavaScript/TypeScript (fal.ai)

import { fal } from '@fal-ai/client'

fal.config({ credentials: 'YOUR_FAL_KEY' })

const result = await fal.subscribe('fal-ai/flux/schnell', {
  input: {
    prompt: "A serene mountain landscape at sunset",
    num_inference_steps: 4,
    image_size: "landscape_16_9"
  }
})

Installation

# Python (Diffusers)
pip install -U diffusers torch

# JavaScript (fal.ai)
npm install --save @fal-ai/client

Model Capabilities

Capability Supported
Text-to-Image Yes
Batch Generation Yes (up to 4 images)
Reproducible Seeds Yes
Custom Aspect Ratios Yes
Multiple Output Formats Yes (WebP, JPG, PNG)
CPU Offloading Yes
Safety Filtering Yes (optional)

Prohibited Uses

  • Violations of applicable laws
  • Exploitation or harm to minors
  • Generation of false information intended to harm
  • Creation of non-consensual nudity or illegal content
  • Harassment, abuse, stalking, or bullying
  • Automated decision-making affecting legal rights
  • Large-scale disinformation campaigns

Availability

Platform URL
Hugging Face https://huggingface.co/black-forest-labs/FLUX.1-schnell
GitHub https://github.com/black-forest-labs/flux
Replicate https://replicate.com/black-forest-labs/flux-schnell
fal.ai https://fal.ai/models/fal-ai/flux/schnell
Mystic.ai https://www.mystic.ai/black-forest-labs/flux1-schnell
BFL Playground https://playground.bfl.ai
ComfyUI Node-based local workflow
Cloudflare Workers AI Hosted inference

Community Statistics (Hugging Face)

  • Downloads: 702,738+ last month
  • Likes: 4,470+
  • Community Discussions: 196+
  • Spaces Using Model: 100+
  • Adapter Models: 261
  • Fine-tunes: 57
  • Quantizations: 27

Company Information

Black Forest Labs is a frontier AI research laboratory focused on visual intelligence, headquartered in Germany. The company recently completed a $300 million Series B funding round.

References

Last Updated: December 2025