HunyuanImage-3-NF4-v2

Hunyuan Image 3.0 Base -- NF4 Quantized (v2)

NF4 (4-bit) quantization of the HunyuanImage-3.0 base model (v2). Fits on a single 48GB GPU. High-quality text-to-image generation with the Hunyuan 3.0 MoE architecture. CFG-distilled for single-pass inference.

What's New in v2

v2 uses improved quantization with more precise skip-module selection, keeping attention projections and critical embedding layers in full BF16 precision for better image quality.

Key Features

• Text-to-image generation with the Hunyuan 3.0 MoE architecture
• NF4 quantized -- ~47 GB on disk
• 45 diffusion steps (CFG-distilled, single-pass)
• Block swap support -- offload transformer blocks to CPU for lower VRAM
• ComfyUI ready -- works with Comfy_HunyuanImage3 nodes

VRAM Requirements

Recommended Hardware:

• Single 48GB GPU (RTX 6000 Ada, RTX PRO 5000, A6000)
• With block swap: may work on 24GB GPUs (swapping ~20 blocks)

Model Details

• Architecture: HunyuanImage-3.0 Mixture-of-Experts Diffusion Transformer
• Parameters: 80B total, 13B active per token (top-K MoE routing)
• Variant: Base (text-to-image)
• Quantization: 4-bit NormalFloat (NF4) quantization via bitsandbytes with double quantization
• Diffusion Steps: 45
• Default Guidance Scale: 7.0
• Resolution: Up to 2048x2048
• Language: English and Chinese prompts

Quantization Details

Layers quantized to NF4:

• Feed-forward networks (FFN/MLP layers)
• Expert layers in MoE architecture (64 experts per layer)
• Large linear transformations

Kept in full precision (BF16):

• VAE encoder/decoder (critical for image quality)
• Attention projection layers (q_proj, k_proj, v_proj, o_proj)
• Patch embedding layers
• Time embedding layers
• Vision model (SigLIP2)
• Final output layers

Usage

ComfyUI (Recommended)

This model is designed to work with the Comfy_HunyuanImage3 custom nodes:

cd ComfyUI/custom_nodes
git clone https://github.com/EricRollei/Comfy_HunyuanImage3

1. Download this model to your preferred models directory
2. Use the "Hunyuan 3 V2 Unified" node
3. Point the model path to this folder and select nf4 precision
4. Set blocks_to_swap to -1 (auto) or a manual value based on your VRAM

Block Swap

Block swap allows running INT8 and BF16 models on GPUs with less VRAM than the full model requires. The system keeps N transformer blocks on CPU and swaps them to GPU on demand during each diffusion step.

Original Model

This is a quantized derivative of Tencent's HunyuanImage-3.0.

• License: Tencent Hunyuan Community License

Credits

• Original Model: Tencent Hunyuan Team
• Quantization: Eric Rollei
• ComfyUI Integration: Comfy_HunyuanImage3

License

This model inherits the license from the original Hunyuan Image 3.0 model: Tencent Hunyuan Community License