Nunchaku R32 SDXL Series: Ultra-Fast 4-bit Quantization
This repository provides a collection of lightweight quantized SDXL models optimized using the Nunchaku (SVDQ W4A4) engine.
Each model in this series is quantized with Rank 32 (r32). While standard quantization keeps higher ranks for quality, r32 is used here to ensure maximum speed and accessibility. This is particularly crucial for:
- Low VRAM Environments: Running SDXL on laptops and GPUs with 6GB-8GB VRAM.
- High-Speed Inference: significantly reducing generation time for rapid prototyping and batch processing.
- Efficiency: Maintaining the core aesthetic of the base model while minimizing the computational footprint.
🚀 Key Features
- Engine: Nunchaku SVDQ (Smooth Vertical-Diagonal Quantization)
- Precision: FP4/NVFP4 (4-bit Weights / 4-bit Activations)
- Rank: 32 (r32) - Aggressive compression for peak performance.
- VRAM Optimized: Fits comfortably in 6GB-8GB VRAM environments where standard SDXL struggles.
- Performance: Native acceleration on NVIDIA RTX 30/40/50 series GPUs.
📦 Available Models
| Filename | Base Model | Version | License |
|---|---|---|---|
realvisxlV50_v50_r32_svdq_fp4.safetensors |
RealVisXL V5.0 | v50.0 | CreativeML Open RAIL++-M |
waiRealCN_v10_r32_svdq_fp4.safetensors |
wai-RealCN | v10.0 | CreativeML Open RAIL++-M |
bluepencilXL_v031_r32_svdq_fp4.safetensors |
BluePencil-XL | v0.3.1 | CreativeML Open RAIL++-M |
waiIllustriousSDXL_v160_r32_svdq_fp4.safetensors |
waiIllustriousSDXL | v1.6.0 | CreativeML Open RAIL++-M |
koronemixIllustrious_v10_r32_svdq_fp4.safetensors |
koronemix-illustrious | v10.0 | CreativeML Open RAIL++-M |
novaanimeXL_v15_r32_svdq_fp4.safetensors |
Nova Anime XL | v15.0 | CreativeML Open RAIL++-M |
🛠 Usage (ComfyUI)
To use these models with full features (Dual CLIP loading, LoRA support, and ControlNet compatibility), you need the Unofficial Nunchaku Loader nodes.
1. Required Custom Nodes
- Nunchaku DiT & LoRA Loader (by ussoewwin): ComfyUI-nunchaku-unofficial-loader
Note: This loader is specifically designed to handle SVDQ-patched UNet/DiT models and provides seamless LoRA integration.
2. Setup
- UNet: Place the
.safetensorsfiles inComfyUI/models/diffusion_models/ - VAE: Use standard SDXL VAE (place in
models/vae/)
📜 Credits & License
🏆 Special Acknowledgement
We extend our deepest respect and gratitude to the Nunchaku Team for their groundbreaking work on SVDQ quantization and for sharing their models with the community. This collection relies heavily on their research and original implementation.
- Original Repository: nunchaku-tech/nunchaku-sdxl
Base Models
These models are derivatives of their respective creators. All credit for aesthetic tuning and model training belongs to the original creators.
- RealVisXL V5.0: Created by SG_161222.
- wai-RealCN: Created by wai.
- BluePencil-XL v0.3.1: Created by blue_pen.
- waiIllustriousSDXL: Created by wai.
- koronemix-illustrious: Created by korone.
- Nova Anime XL: Created by realdos.
Software & Integration
- ComfyUI Loaders: The Nunchaku SDXL DiT Loader and LoRA Loader were developed and are maintained