I’ve been experimenting with the new Qwen Image 2.1 release and made an NVFP4 quant of its diffusion model. It’s working well in my local tests, so I’m sharing it for non-commercial research and evaluation. Hope it helps others trying the release.Built with Qwen. Original model by the <a target="_blank" rel="ugc" href="https://huggingface.co/Qwen/Qwen-Image-2.1">Qwen team</a>. Here are the direct downloads for the Qwen Image 2.1 these are separate downloads. <ul><li><a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/text_encoders/qwen3vl_8b_w4a8.safetensors?download=true">Text encoder — Qwen3VL 8B W4A8 (6.31 GB)</a></li><li><a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/vae/qwen_image_2.1_vae_bf16.safetensors?download=true">VAE — BF16 (676 MB)</a> </li><li>See the <a target="_blank" rel="ugc" href="https://github.com/QwenLM/Qwen-Image-2.1/blob/main/LICENSE">Qwen Research License</a>.</li></ul>Starting point: Euler, 40 steps, CFG 1; Normal scheduler❤️🙏

Files with SHA256: 97ea02fad5deab9297995453f588a05da183a1136deca911e8fb09dd6ce8a3f0