EricRollei/HunyuanImage-3.0-Instruct-Distil-NF4

Hunyuan Image 3.0 Instruct Distil — NF4 Quantized

NF4 (4-bit) quantization of the HunyuanImage-3.0 Instruct Distil model. The most accessible option — fits on a single 48GB GPU with ~6x faster generation (8 steps vs 50). Best balance of speed, quality, and VRAM.

Key Features

• 🎯 Instruct model — supports text-to-image, image editing, multi-image fusion
• 🧠 Chain-of-Thought — built-in think_recaption mode for highest quality
• 💾 NF4 quantized — ~45 GB on disk
• ⚡ 8 diffusion steps (CFG-distilled for speed)
• 🔧 ComfyUI ready — works with Comfy_HunyuanImage3 nodes

VRAM Requirements

Component	Memory
Weight Loading	~29 GB weights
Inference (additional)	~12-20 GB inference
Total	~41-49 GB

Recommended Hardware:

• Fits on a single 48GB GPU (RTX 6000 Ada, RTX PRO 5000, A6000)
• Consumer GPUs (RTX 4090/5090 24GB) — not enough VRAM

Model Details

• Architecture: HunyuanImage-3.0 Mixture-of-Experts Diffusion Transformer
• Parameters: 80B total, 13B active per token (top-K MoE routing)
• Variant: Instruct Distil (CFG-Distilled, 8-step)
• Quantization: 4-bit NormalFloat (NF4) quantization via bitsandbytes with double quantization
• Diffusion Steps: 8
• Default Guidance Scale: 2.5
• Resolution: Up to 2048x2048
• Language: English and Chinese prompts

Distillation

This is the CFG-Distilled variant, which means:

• Only 8 diffusion steps needed (vs 50 for the full Instruct model)
• ~6x faster image generation
• No quality loss — distilled to match the full model's output
• cfg_distilled: true in config means no classifier-free guidance needed

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 ComfyUI models directory
2. Use the "Hunyuan 3 Instruct Loader" node
3. Select this model folder and choose nf4 precision
4. Connect to the "Hunyuan 3 Instruct Generate" node for text-to-image
5. Or use "Hunyuan 3 Instruct Edit" for image editing
6. Or use "Hunyuan 3 Instruct Multi-Fusion" for combining multiple images

Bot Task Modes

The Instruct model supports three generation modes:

Mode	Description	Speed
image	Direct text-to-image, prompt used as-is	Fastest
recaption	Model rewrites prompt into detailed description, then generates	Medium
think_recaption	CoT reasoning → prompt enhancement → generation (best quality)	Slowest

Original Model

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

• Architecture: Diffusion Transformer with Mixture-of-Experts
• Resolution: Up to 2048x2048
• Language Support: English and Chinese prompts
• License: Tencent Hunyuan Community License

Limitations

• Requires high-end professional GPU (~41-49 GB VRAM)
• NF4 quantization may introduce minor quality differences in edge cases
• Loading time adds ~1-2 minutes overhead to first generation
• CoT/recaption modes require additional time for text generation phase

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

Please review the original license for commercial use restrictions and requirements.

Citation

@misc{hunyuan-image-3-nf4-instruct,
  author = {Rollei, Eric},
  title = {Hunyuan Image 3.0 Instruct Distil — NF4 Quantized},
  year = {2026},
  publisher = {Hugging Face},
  howpublished = {\url{https://huggingface.co/EricRollei/HunyuanImage-3.0-Instruct-Distil-NF4}}
}