← sources

New few-step LoRA adapters for Qwen-Image-2.1, ~5x faster

source: reddit (reddit:StableDiffusion)published 2026-09-24 16:38ingested 2026-09-25 02:40

retained text

Hey! We published a set of a few-step LoRAs: [https://huggingface.co/PrunaAI/Pruna-Qwen-Image-2.1](https://huggingface.co/PrunaAI/Pruna-Qwen-Image-2.1)

\- Generate or edit images in just 5 or 8 steps instead of 40. Choose 5 steps for maximum speed or 8 steps for the best balance.  
\- The 8-step adapter is our recommended default. It supports text-to-image and single- or multi-image editing at 1K resolution, with up to three reference images.  
\- The LoRA adapters load directly on top of Qwen-Image-2.1, while the pipeline remains unchanged. Each adapter uses its own optimized sigma schedule and runs without CFG.

document fetched 2026-09-25 00:33 · canonical https://reddit.com/r/StableDiffusion/comments/1wp6la7/new_fewstep_lora_adapters_for_qwenimage21_5x