self-hosted/ai
§01·model · /models

Qwen-Image-2.1

imageactiveQwen Research License (non-commercial)

Alibaba Qwen's Qwen-Image-2.1 — a 7B single-stream diffusion transformer (32 layers) behind a Qwen3-VL-8B text encoder and a new 64-channel RGBA autoencoder, released 2026-09-20. One checkpoint covers text-to-image and instruction-based editing with up to ten reference images, renders natively at 2K (2048×2048) and can output a real alpha channel. It is far smaller than the 20B Qwen-Image it follows: the transformer is 13.25 GiB in bf16 and 6.76 GiB in Comfy-Org's int8 repack, so on consumer cards the text encoder (16.33 GiB bf16, 8.71 GiB int8, 5.88 GiB w4a8) is the larger file. The licence changed with the size: the Qwen Research License Agreement permits non-commercial use only, where Qwen-Image was Apache-2.0. Native in ComfyUI from v0.37.0.

§02·same family
1 other model

Other models grouped with Qwen-Image-2.1. Sizes and modalities can differ, and nothing here says whether one fits your GPU — open a model for its own compatibility table.

§03·GPUs that run this model
4 total

benchmarked·~ runs via recipe (not benchmarked)· untested·doesn't fit