Qwen releases 8-step accelerated Qwen-Image-2.1-Turbo on Hugging Face
multimodalart · x · 2026-10-09
Alibaba's Qwen team has released Qwen-Image-2.1-Turbo on Hugging Face, an accelerated checkpoint of Qwen-Image-2.1 for text-to-image generation and image editing that cuts denoising to just 8 steps while keeping the same 7B visual generation architecture.
The checkpoint ships with its recommended sampling schedule, so it works out of the box without manual scheduler configuration. It defaults to CFG=1 and uses prefix KV caching to reuse text and reference-image context across denoising steps. It loads directly via the QwenImage21Pipeline in Diffusers, though it requires the latest Diffusers source with sampling-sigma support from PR #14950.
Related event: Alibaba Open-Sources Qwen-Image-2.1-Turbo: 2K Images in 8 Denoising Steps(11 posts)→
More from Multimodal
- MIRA agent refines musical intent, lifting open-source music gen to Suno-level — Zekai Liu · 2026-10-10
- Magnific praised for releasing a new model claimed better than Midjourney — cuenca · 2026-10-10
- Qwen-Image-2.1-Turbo gets official ComfyUI support on Hugging Face — Time-Teaching1926 · 2026-10-09
- 'Fly through your wallpaper': image-to-video prompt goes viral — umesh_ai · 2026-10-09
- PoolDINO cuts RAE image generation tokens 4-16x, runs on an M1 Pro CPU — francoisfleuret · 2026-10-09
- Creator tests Kling 4.0 cut against Seedance 2.5 in action short experiment — azed_ai · 2026-10-09