Skip to page content
SAVRN
Search Plan a project

Open-weight model · Text to image

MGFlow

by Haoyang Shi shy0423/MGFlow

MGFlow is an open-weight model for text to image from Haoyang Shi. Its published files total 76.0 GB.

Model checkpoints, reference distributions, and evaluation assets for MGFlow, presented in Unifying Distributional Training for One-Step Visual Generation.

Parameters—
Context—
Weights69.7 GB
License—
AccessOpen weights
Monthly Downloads—

Model Card

Model checkpoints, reference distributions, and evaluation assets for MGFlow, presented in Unifying Distributional Training for One-Step Visual Generation. The release covers ImageNet class-conditional generation at 256×256 with JiT and pMF (B/L/H), and text-to-image generation at 512×512 with FLUX.2 [klein] 4B. Both generate images in one step after MGFlow post-training. The T2I checkpoints use the MGFlow-T2I reference dataset, which contains synthetic COCO and GenEval images paired with their prompts. The dataset link in the metadata applies to this T2I branch. If you use MGFlow or these assets, please cite

Excerpt from the card by Haoyang Shi.

Identity and Version

Repository
shy0423/MGFlow
Publisher
Haoyang Shi
Task
Text to image
Modality
Image
Library
Not stated by the source
Parameters
Not stated by the source
Languages
Not stated by the source
Revision
222ebaf8711521ec0aba7c1b0a22631f28d61da4
First published
2026-09-29
Last updated
2026-10-02

Files and Weights

56 files, 76.0 GB in total. The weights are 48 files totalling 69.7 GB in npz, pth.

Weights48 files · 69.7 GB
Documentation4 files · 4.0 KB
Other3 files · 6.3 GB
Repository1 file · 1.5 KB
Every file
FileTypeSizeSHA-256
Checkpoints/ImageNet/Base/JiT-B.pthWeights525.3 MB e99eb02e5628
Checkpoints/ImageNet/Base/JiT-H.pthWeights3.8 GB fc9e40890605
Checkpoints/ImageNet/Base/JiT-L.pthWeights1.8 GB 7fb543facf70
Checkpoints/ImageNet/Base/pMF-B.pthWeights473.0 MB ef08492ef0ba
Checkpoints/ImageNet/Base/pMF-H.pthWeights3.8 GB 54ba66a103a8
Checkpoints/ImageNet/Base/pMF-L.pthWeights1.6 GB 475010c02de7
Checkpoints/ImageNet/Post-trained/JiT-B_KL-k1+4+16.pthWeights525.3 MB 6b1624f2c505
Checkpoints/ImageNet/Post-trained/JiT-B_W2-k1+4.pthWeights525.3 MB c126de14dd82
Checkpoints/ImageNet/Post-trained/JiT-H_KL-k1+4+16.pthWeights3.8 GB 6b01d66ca04b
Checkpoints/ImageNet/Post-trained/JiT-H_W2-k1+4.pthWeights3.8 GB b9b7f3a20204
Checkpoints/ImageNet/Post-trained/JiT-L_KL-k1+4+16.pthWeights1.8 GB aa0f97d4b106
Checkpoints/ImageNet/Post-trained/JiT-L_W2-k1+4.pthWeights1.8 GB e15bd834df14
Checkpoints/ImageNet/Post-trained/pMF-B_KL-k1+4+16.pthWeights473.0 MB a74aa31a17c9
Checkpoints/ImageNet/Post-trained/pMF-B_W2-k1+4.pthWeights473.0 MB b8257939c068
Checkpoints/ImageNet/Post-trained/pMF-H_KL-k1+4+16.pthWeights3.8 GB 832ba0aade01
Checkpoints/ImageNet/Post-trained/pMF-H_W2-k1+4.pthWeights3.8 GB e3201902a946
Checkpoints/ImageNet/Post-trained/pMF-L_KL-k1+4+16.pthWeights1.6 GB 622e5834728f
Checkpoints/ImageNet/Post-trained/pMF-L_W2-k1+4.pthWeights1.6 GB 6d969fb55862
Checkpoints/T2I/FLUX2-klein-4B_COCO-image.pthWeights15.5 GB 99b35e98ff81
Checkpoints/T2I/FLUX2-klein-4B_COCO-joint.pthWeights15.5 GB 8577d1e3c7e2
Evaluation/ImageNet/stats/convnext_in256_t224_stats.npzWeights8.4 MB 5b9be5c8425c
Evaluation/ImageNet/stats/guided_diffusion_stats.npzWeights33.6 MB f70a7b30c862
Evaluation/ImageNet/stats/jit_in256_stats.npzWeights33.6 MB 925e8b5b4ced
Evaluation/ImageNet/stats/vit_large_patch14_clip_224_openai_in256_t256_stats.npzWeights16.8 MB e13f614d1844
Evaluation/ImageNet/stats/vit_large_patch14_dinov2_lvd142m_in256_t256_stats.npzWeights16.8 MB fca8c89fe1c8
Evaluation/ImageNet/stats/vit_large_patch16_224_mae_in256_t224_stats.npzWeights16.8 MB f9f2c09cf9fc
Evaluation/ImageNet/stats/vit_so400m_patch16_siglip_256_v2_webli_in256_t224_stats.npzWeights21.3 MB a0d16403d390
Reference/ImageNet/Inception_K1.npzWeights33.6 MB b360adc4539b
Reference/ImageNet/Inception_K16.npzWeights537.1 MB 0bf6b6ab6f2a
Reference/ImageNet/Inception_K4.npzWeights134.3 MB 4257c15f531c
Reference/ImageNet/MAE_K1.npzWeights8.4 MB 3666df721452
Reference/ImageNet/MAE_K16.npzWeights134.3 MB cc0ece2c5e02
Reference/ImageNet/MAE_K4.npzWeights33.6 MB 93b1a2d402a2
Reference/ImageNet/SigLIP_K1.npzWeights10.6 MB 7a819070af18
Reference/ImageNet/SigLIP_K16.npzWeights170.0 MB e1cd8a0e0d21
Reference/ImageNet/SigLIP_K4.npzWeights42.5 MB 8f5ddae41665
Reference/T2I/COCO-image/Inception_K1.npzWeights33.6 MB 6586be259151
Reference/T2I/COCO-image/Inception_K4.npzWeights134.3 MB 3b35af865b26
Reference/T2I/COCO-image/MAE_K1.npzWeights8.4 MB 6529c62a19d5
Reference/T2I/COCO-image/MAE_K4.npzWeights33.6 MB dda5d4bde6a2
Reference/T2I/COCO-image/SigLIP_K1.npzWeights10.6 MB 784353d3e7d4
Reference/T2I/COCO-image/SigLIP_K4.npzWeights42.5 MB bf0bc6f410eb
Reference/T2I/COCO-joint/Inception_K1.npzWeights81.9 MB 6dbe8aad0282
Reference/T2I/COCO-joint/Inception_K4.npzWeights327.8 MB 139e3d472384
Reference/T2I/COCO-joint/MAE_K1.npzWeights37.9 MB 94b78f1823d3
Reference/T2I/COCO-joint/MAE_K4.npzWeights151.6 MB 703d93014ee0
Reference/T2I/COCO-joint/SigLIP_K1.npzWeights42.5 MB 9e31f45577d3
Reference/T2I/COCO-joint/SigLIP_K4.npzWeights169.9 MB bed5ebe515ae
Evaluation/LICENSE.txtDocumentation1.1 KB —
README.mdDocumentation2.1 KB —
Reference/README.mdDocumentation486 B —
Reference/T2I/README.mdDocumentation325 B —
Checkpoints/LICENSES/JiT.txtOther1.1 KB —
Checkpoints/LICENSES/pMF.txtOther1.1 KB —
Reference/Encoders.tarOther6.3 GB 837146ab51e9
.gitattributesRepository1.5 KB —

License and Download

License
Not stated by the source
Access
Open weights, no gate
Download size
69.7 GB
Download from Haoyang Shi

Released by Haoyang Shi through its official repository on Hugging Face.

Built From

Memory Requirements

PrecisionWeights in memory
As published69.7 GB

Weights only, from the published parameter count; the key-value cache and runtime add to this.

Similar Models

GGUF, Safetensors, and MLX quantized formats of Qwen/Qwen-Image-2.1 for local image generation using the original upstream base weights. Q4KM is recommended for the best balance of size and quality for GGUF. INT8 ConvRot and NVFP4 offer high performance on supported GPUs, while MLX formats are optimized for Apple Silicon. Companion model files packaged for ComfyUI: Use the model with ComfyUI and ComfyUI-GGUF. All required companion files (GGUF transformer, text encoder, and VAE) are hosted directly in this repository. Download the files and place them in their respective ComfyUI directories: 1. Install ComfyUI-GGUF: Use the maintained fork with native Qwen-Image 2.1 support by cloning…

Open weights other gguf
View model

GGUF quantizations of Qwen/Qwen-Image-2.1 for local image generation using the original upstream base weights. Q4KM is recommended for the best balance of size and quality. Companion model files packaged for ComfyUI: Use the model with ComfyUI and ComfyUI-GGUF. All required companion files (GGUF transformer, text encoder, and VAE) are hosted directly in this repository. Download the files and place them in their respective ComfyUI directories: 1. Install ComfyUI-GGUF: Use the maintained fork with native Qwen-Image 2.1 support by cloning leejet/ComfyUI-GGUF into your custom nodes: (Note: If you have the older city96/ComfyUI-GGUF installed and encounter an Unknown model architecture! error…

Open weights other gguf
View model

Model · Text to image

Qwen-Image-2.1-GGUF

Unsloth AI

This is a GGUF quantized version of Qwen-Image-2.1. unsloth/Qwen-Image-2.1-GGUF uses Unsloth Dynamic 2.0 methodology for SOTA performance. - Important layers are upcasted to higher precision, per tensor, from a measured sensitivity scan. - Run these with Unsloth Desktop, stable-diffusion.cpp and more. A GGUF is the denoiser only, so it needs the VAE and the Qwen3-VL text encoder alongside it. See below for image editing operating inside of Unsloth Desktop: Rendered with the Q4KM denoiser and the Q4KM text encoder, 1024x1024, 20 steps, cfg 6.0, euler. We are excited to open-source Qwen-Image-2.1, a unified text-to-image generation and image editing model in the Qwen family. With just 7B…

Open weights other
View model

Model · Text to image

Z-Image-Turbo-GGUF

Unsloth AI

Welcome to the official repository for the Z-Image（造相）project! Z-Image is a powerful and highly efficient image generation model with 6B parameters. Currently there are three variants: - Z-Image-Turbo – A distilled version of Z-Image that matches or exceeds leading competitors with only 8 NFEs (Number of Function Evaluations). It offers sub-second inference latency on enterprise-grade H800 GPUs and fits comfortably within 16G VRAM consumer devices. It excels in photorealistic image generation, bilingual text rendering (English & Chinese), and robust instruction adherence. - Z-Image-Base – The non-distilled foundation model. By releasing this checkpoint, we aim to unlock the full potential…

Open weights apache-2.0 ggml
View model

Model · Text to image

Juggernaut-XL-v9

RunDiffusion

The SDXL ecosystem is the single most mature corner of open image generation, and v9 is its most refined photorealism checkpoint. Choose Juggernaut XL v9 when you want: - Photorealism that holds up under scrutiny — skin texture, micro-contrast, and natural lighting that translates from concept to print. - Reasonable hardware — runs comfortably on 8 GB of VRAM, unlike newer DiT-based models that demand 16+ GB. - The full SDXL toolbox — drop-in compatibility with the thousands of SDXL ControlNets, IP-Adapter variants, AnimateDiff, regional prompting tools, and LoRAs already in your workflow. - Battle-tested reliability — 26+ months in production, used in agencies, studios, and shipping…

Open weights creativeml-openrail-m diffusers
View model

Model · Text to image

Qwen-Image-Lightning

Lightx2v

Please refer to Qwen-Image-Lightning github to learn how to use the models. make sure to install diffusers from main (pip install git+https://github.com/huggingface/diffusers.git)

Open weights apache-2.0 diffusers
View model