Downloads · 30 days
1K
39% of all-time downloads
mflux-community/ernie-image-base-mflux-q4
ernie-image-base-mflux-q4 is a text-to-image model from mflux-community. Use it when you need an image from a text prompt. It is set up for mflux. The card lists the license as apache-2.0.
This is a Q4 MFlux version of ERNIE-Image.
Downloads · 30 days
1K
39% of all-time downloads
All-time downloads
2.6K
Public
Repo size
6.6 GB
Likes
0
Public
Click a slice to open those files.
.safetensors6.6 GB · 100%
From the Hugging Face model README
This is a Q4 MFlux version of ERNIE-Image.
MFlux runs the latest state-of-the-art generative image models locally on your Mac in native MLX. MFlux is open source and free.
[!NOTE] Community conversion by MFlux-Community -- not an official release from the original model's authors. The original model's license applies unchanged, see License.
| QUANT | GB | Repo |
|---|---|---|
| BF16 | 23.1 | mflux-community/ernie-image-base-mflux-bf16 |
| Q8 | 12.4 | mflux-community/ernie-image-base-mflux-q8 |
| Q6 | 9.5 | mflux-community/ernie-image-base-mflux-q6 |
| Q5 | 8.1 | mflux-community/ernie-image-base-mflux-q5 |
| Q4 | 6.6 | this repo |
| Q3 | 5.2 | mflux-community/ernie-image-base-mflux-q3 |
Converted on 2026-08-04.
Standard MFlux quantization uses MLX affine quantization (4-bit) across the text encoder, transformer and VAE. The quantization level is stored in the weights, so no --quantize flag is needed.
Install MFlux
uv tool install --upgrade mflux
Generate an image using ERNIE-Image Q4:
mflux-generate-ernie-image \
--model mflux-community/ernie-image-base-mflux-q4 \
--base-model ernie-image \
--prompt "A puffin standing on a cliff" \
--width 1024 \
--height 1024 \
--seed 42 \
--output image.png
The weights are downloaded from this repo on first use. To keep a local copy instead:
hf download mflux-community/ernie-image-base-mflux-q4 --local-dir ./ernie-image-base-mflux-q4
then pass --model ./ernie-image-base-mflux-q4.
This repo inherits the license of the original model: apache-2.0. See the original model card for the full terms and any usage restrictions. The weights here were modified from the original release only by conversion to MLX and quantization.