SDXL Turbo
Parameters
2.56B
Modality
Image
Auxiliary Parameters
890M
Architecture
Diffusion Transformer
License
Stability AI Community License
Release Date
28 Nov 2023
API Pricing (per 1M)
Self-hosted only
Hardware Sizing
Precision
Workload Memory: 3.74 GB
INT4 · 1024x1024
DiT Backbone Weights: 1.19 GB
VAE Decode Workspace: 0.05 GB
Runtime & Graphs: 2.50 GB
Minimum Required Hardware
1x NVIDIA L4 24GB
3.7 / 21.6 GB usable
1x RTX 3060 12GB
3.7 / 10.8 GB usable
Apple M4 24GB
3.7 / 18.0 GB usable
Determine requirements for your specific workload:
Open in VRAM Calculator
Architecture Diagram
About SDXL Turbo
SDXL Turbo is a real-time single-step diffusion model trained with Adversarial Diffusion Distillation (ADD), generating 512x512 outputs in a single forward pass with low inference latency.
Technical Specifications
Attention
Attention Heads
102
Attention Head Dimension
20
Dimensions
Auxiliary Parameters
890M
Hidden Dimension Size
2,048
Number of Layers
6