<p>Ready-to-use ComfyUI workflow for <strong>Qwen-Image 2.1 Turbo</strong>, Qwen's official 8-step model.</p><h3 id="what-you-need">What you need</h3><ul><li><p><strong>Turbo model:</strong> qwen_image_2.1_turbo_int8_convrot.safetensors (7.26 GB) in models/diffusion_models</p></li><li><p><strong>Text encoder:</strong> qwen3vl_8b_int8_convrot.safetensors in models/text_encoders</p></li><li><p><strong>VAE:</strong> qwen_image_2.1_vae_bf16.safetensors in models/vae</p></li><li><p><strong>ComfyUI</strong> 0.37.0 or newer</p></li></ul><p>All three files are on Hugging Face in Comfy-Org/Qwen-Image-2.1. If you already use Qwen-Image 2.1, you only need the Turbo model.</p><h3 id="how-to-use">How to use</h3><ol><li><p>Drag the workflow into ComfyUI.</p></li><li><p>Write your prompt and pick a size.</p></li><li><p>Run</p></li></ol><h3 id="settings">Settings</h3><p>Everything is already set: <strong>8 steps, cfg 1, euler</strong>, with the sigma schedule Turbo was trained on (in a ManualSigmas node). Leave these as they are. The negative prompt has no effect at cfg 1.</p><h3 id="speed">Speed</h3><p>Time per 1024 x 1024 image, Turbo vs the regular Qwen-Image 2.1 (25 steps):</p><ul><li><p><strong>RTX 3060 12 GB</strong> : <strong>15.0 s (Turbo)</strong> vs 34.9 s (Base)</p></li><li><p><strong>RTX 5090</strong> <strong>32 GB</strong>: <strong>2.0 s (Turbo)</strong> vs 4.5 s (Base)</p></li></ul><p>See related video that compares the performance.</p><div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/QiSVLf-CGDI" start="0"></iframe></div><h3 id="licence">Licence</h3><p>Qwen Research Licence, <strong>non-commercial</strong>, the same as Qwen-Image 2.1.</p><p></p>

Qwen-Image 2.1 Turbo · 8 steps · ComfyUI workflow

<p>First release. ComfyUI's Qwen-Image 2.1 template with the Turbo model and Turbo's trained 8-step schedule (ManualSigmas).</p>

Files with SHA256: 50b9d37c04076d06c9296fd6dcf588162a69eeb5101a58b214681930063e7d9b