Qwen is licensed under the Qwen RESEARCH LICENSE AGREEMENT, Copyright (c) 2026 Hangzhou Tongyi
Laboratory Technology Co., Ltd. All Rights Reserved.

This repository distributes a MODIFIED form of Qwen-Image 2.1. The weights were converted for
libwaifu (https://github.com/ling0322/libwaifu): the tensors were renamed and refused into the
layout its runtime reads (the attention projections fused into one matrix each), the denoiser's
zero-centred RMSNorm was folded to a plain multiplier, and everything was narrowed from bfloat16
to float16 (and, in the fp8 package, the matrices quantized to E4M3 with one scale per output
channel). No weight was retrained or fine-tuned.

This project is Built with Qwen. It is not an official Qwen product and is not endorsed by
Alibaba or the Qwen team.

Recipients of these files are bound by the Qwen RESEARCH LICENSE AGREEMENT, a copy of which is
distributed alongside them as LICENSE. The agreement grants non-commercial (research and
evaluation) use only; see LICENSE for the exact terms.