---
license: apache-2.0
base_model: Qwen/Qwen-Image-2.1
base_model_relation: quantized
library_name: mlx-serve
tags:
  - mlx
  - mlx-serve
  - quantized
  - text-to-image
  - image-to-image
pipeline_tag: text-to-image
---

# Qwen-Image-2.1 MLX-Serve 4-bit

4-bit pack of [Qwen/Qwen-Image-2.1](https://huggingface.co/Qwen/Qwen-Image-2.1) for
[mlx-serve](https://github.com/ddalcu/mlx-serve): 11.2 GB, for 16 GB Macs.


![sample](https://raw.githubusercontent.com/ddalcu/mlx-serve/main/website/screenshots/qwen-image-2.1-4bit-512.jpg)

## What is in it

The checkpoint's own diffusers layout and key names, with the DiT block linears and the
text-encoder layer linears affine-quantized to 4-bit (group 64). Kept dense: the VAE
(f32), `embed_tokens`, norms, and the DiT's small or shared linears. Kept: the Qwen3-VL
vision tower, for instruction editing. Dropped: `lm_head` and the VAE's per-frame `time_conv`s.
Built by `tests/convert_qwen_image21_weights.py --preset 16gb`.

## Measured (M1 Pro, 32 GB)

| Pack | Size | Steps | Wall clock incl. load | Peak memory |
|---|---|---|---|---|
| 8-bit | 1024x1024 | 40 | 985 s (~23 s/step) | 12.95 GB |
| 4-bit | 1024x1024 | 3 | 87 s | 9.55 GB |
| 4-bit | 512x512 | 20 | 118 s | - |

On a Mac the full set would crowd, mlx-serve loads the text encoder per request and frees
it before the denoise, so the resident set is the DiT and VAE.

## Run it

```sh
brew tap ddalcu/mlx-serve https://github.com/ddalcu/mlx-serve
brew install mlx-serve
mlx-serve pull ddalcu/Qwen-Image-2.1-MLX-Serve-4bit
mlx-serve serve
curl localhost:11234/v1/images/generations -H 'Content-Type: application/json' \
  -d '{"model":"ddalcu/Qwen-Image-2.1-MLX-Serve-4bit","prompt":"a red fox in fresh snow","size":"1024x1024"}'
```

40 steps when `steps` is omitted. `guidance_scale` above 1 with a `negative_prompt` runs
real CFG (two forwards per step). `image` + `strength` does image-to-image. `"mode":"edit"` with an `image` (plus up to 9 `ref_images`)
edits it from the prompt.

Apache-2.0, same as the base model.