drbaph/HunyuanImage-2.1_fp8
<div align="center"> <img src="https://cdn-uploads.huggingface.co/production/uploads/63473b59e5c0717e6737b872/5DZez8C7TeFwRn3FcKDix.png" alt="HunyuanImage-2.1 Banner" /> <h1> HunyuanImage-2.1 fp8 e4m3fn </h1> <h2>An Efficient Diffusion Model for High-Resolution (2K) Text-to-Image Generation</h2> </div>
</div> <div align="center"> <a href="https://github.com/Tencent-Hunyuan/HunyuanImage-2.1" target="blank"><img src="https://img.shields.io/badge/Code-black.svg?logo=github" height="22px"></a> <a href="https://huggingface.co/spaces/tencent/HunyuanImage-2.1" target="blank"> <img src="https://img.shields.io/badge/Demo%20Page-blue" height="22px"></a> <a href="https://huggingface.co/tencent/HunyuanImage-2.1" target="blank"><img src="https://img.shields.io/badge/%F0%9F%A4%97%20Models-d96902.svg" height="22px"></a> <a href="#" target="blank"><img src="https://img.shields.io/badge/Report-Coming%20Soon-blue" height="22px"></a> <a href="https://hunyuan-promptenhancer.github.io/" target="blank"><img src="https://img.shields.io/badge/PromptEnhancer-bb8a2e.svg?logo=github" height="22px"></a> <a href="https://x.com/TencentHunyuan" target="blank"><img src="https://img.shields.io/badge/Hunyuan-black.svg?logo=x" height="22px"></a> </div>
Performance on RTX 5090
When using HunyuanImage-2.1 with the quantized encoder + quantized base model, the VRAM usage on an NVIDIA RTX 5090 typically ranges between 26 GB and 30 GB with average 16 second inference time depending on resolution, batch size, and prompt complexity. Reports that it works on 16gb VRAM GPU's
⚠ Important Note: The refiner is still not implemented and is not ready for use in ComfyUI. However, the distilled model now works in ComfyUI with recommended settings of 8 steps / 1.5-2.5 CFG.
<p align="center"> <img src="https://cdn-uploads.huggingface.co/production/uploads/63473b59e5c0717e6737b872/auZ_xmiKPw0QdBYUrTLn-.png" alt="Image1"/> </p> <p align="center"> <img src="https://cdn-uploads.huggingface.co/production/uploads/63473b59e5c0717e6737b872/qod1zCPWjzOZSNcOWx49-.png" alt="Image2"/> </p>
---
Download Quantized Model (FP8 e4m3fn)
**Download hunyuanimage2.1_fp8_e4m3fn.safetensors** ---
Workflow Notes
- Model: HunyuanImage-2.1
- Mode: Quantized Encoder + Quantized Base Model
- VRAM Usage: ~26GB–30GB on RTX 5090
- Resolution Tested: 2K (2048×2048)
- Frameworks: ComfyUI & Diffusers
- Optimisations Works with Patch Sage Attention + Lazycache / TeaCache ✅
- Distilled Model: ✅ Now works in ComfyUI with 8 steps / 1.5-2.5 CFG
- Refiner: ❌ Still not implemented, not available in ComfyUI
- License: tencent-hunyuan-community --- <p align="center"> 🚀 Optimized for High-Resolution, Memory-Efficient Text-to-Image Generation </p>