As of: 2026-11-13 13:49 CET.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/
Canonical page: https://postcutoff.com/m/qwen-image-2-1/

# Qwen-Image-2.1

Alibaba (Qwen), Qwen-Image family. Status: current. Type: image generation. Released: 2026-09-20.

## How to call it

| Provider | Model id | Endpoint | Docs |
|---|---|---|---|
| Hugging Face | `Qwen/Qwen-Image-2.1` |  | https://huggingface.co/Qwen/Qwen-Image-2.1 |
| ModelScope |  |  | https://www.modelscope.cn/models/Qwen/Qwen-Image-2.1 |
| Hugging Face (Turbo, 8-step) | `Qwen/Qwen-Image-2.1-Turbo` |  | https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo |
| GitHub |  |  | https://github.com/QwenLM/Qwen-Image-2.1 |
| Hugging Face Space (demo) |  |  | https://huggingface.co/spaces/Qwen/Qwen-Image-2.1 |

## Pricing

- Price not published.
- Last checked: 2026-10-10

## Specs

- Input: text, image
- Output: image
- Context window: not published
- Max output: not published
- Open weights: yes
- Licence: qwen-research

## Where it stands

- Newer image generation models in the registry since its release: 6. The 5 newest:
  - GPT Image 2.6 Flare (2026-11-04) https://postcutoff.com/m/gpt-image-2-6-flare/
  - GPT Image 2.6 Sunburst (2026-11-04) https://postcutoff.com/m/gpt-image-2-6-sunburst/
  - MOSS-Image (2026-10-16) https://postcutoff.com/m/moss-image/
  - FLUX 3 Image Klein (2026-10-15) https://postcutoff.com/m/flux-3-image-klein/
  - Krea 3 (Fast / High / Ultra modes) (2026-10-15) https://postcutoff.com/m/krea-3/
- Before it from Alibaba: Qwen-Image-3.0 (Pro) (2026-07-21) https://postcutoff.com/m/qwen-image-3-0/
- Weights are public. Licence: Qwen Research License. See the licence for commercial terms.

## What stands out

- Native RGBA generation and editing: Generates transparent images, edits transparent layers and extracts subjects from photos in one model. Source: https://github.com/QwenLM/Qwen-Image-2.1
- Multi-reference editing: Up to 10 reference images; local edits via circles, painted annotations or masks, with identity preservation. Source: https://github.com/QwenLM/Qwen-Image-2.1

## Notes

7B visual generator (32 single-stream DiT layers). Diffusers pipeline QwenImage21Pipeline (install diffusers from git); day-0 ComfyUI, vLLM-Omni, SGLang support. Qwen Research License, check terms before commercial use. Qwen-Image-2.1-Turbo (HF, 2026-10-09): accelerated checkpoint, 8 denoising steps, CFG=1, same 7B architecture and pipeline, same qwen-research license; schedule saved in the checkpoint (needs Diffusers with pipeline-configured sigmas, PR #14950).

Quick start (from the README):

```python
import torch
from diffusers import QwenImage21Pipeline
pipe = QwenImage21Pipeline.from_pretrained("Qwen/Qwen-Image-2.1", torch_dtype=torch.bfloat16).to("cuda")
image = pipe(prompt="A neon shop sign that reads \"QWEN IMAGE 2.1\"", num_inference_steps=40).images[0]
```

Model page: https://postcutoff.com/m/qwen-image-2-1/