Mage-Flow: An Efficient Native-Resolution Foundation Model for Image Generation and Editing

Hugging Face Daily Papers Papers
image-generation image-editing diffusion-transformer rectified-flow native-resolution efficient microsoft

Summary

Mage-Flow is a compact 4B-parameter generative stack for efficient text-to-image generation and instruction-based image editing, featuring a co-designed lightweight tokenizer (Mage-VAE) and a native-resolution multimodal diffusion transformer trained with rectified flow matching. It achieves competitive performance while enabling high-resolution generation at 0.59s on a single A100 GPU.

Large-scale visual generators are increasingly capable but costly to train, fine-tune, and deploy. We introduce Mage-Flow, a compact 4B-scale generative stack for efficient text-to-image generation and instruction-based image editing. The stack is built from two co-designed components: Mage-VAE, a lightweight high-fidelity latent tokenizer, and a Native-Resolution Multimodal Diffusion Transformer trained with rectified flow matching. Mage-VAE uses one-step diffusion-style encoding and decoding with anchor-latent regularization, preserving the reconstruction quality of strong public VAEs while reducing tokenization cost by more than an order of magnitude. Together with native-resolution packing and stack-level CUDA kernel fusion, the stack supports flexible-resolution training and improves end-to-end training throughput by about 2.5times. Built on this foundation, we develop a complete model family with Base, RL-aligned, and Turbo variants for both generation and editing. Diffusion-NFT improves prompt following, text rendering, aesthetic quality, and editing fidelity, while few-step distillation with adversarial perceptual guidance produces 4-step Turbo models for low-latency inference. Despite its compact scale, Mage-Flow and Mage-Flow-Edit achieves competitive performance across standard generation and editing benchmarks. More importantly, the Turbo variants make high-resolution generation and editing practical for interactive use: at 1024^2 resolution on a single NVIDIA A100 GPU, Mage-Flow-Turbo generates an image in 0.59s, and Mage-Flow-Edit-Turbo edits an image in 1.02s, while maintaining a small memory footprint. These results show that careful tokenizer--backbone--system co-design can deliver strong high-resolution generation and editing within an efficient 4B model family.
Original Article
View Cached Full Text

Cached at: 07/22/26, 06:41 AM

Paper page - Mage-Flow: An Efficient Native-Resolution Foundation Model for Image Generation and Editing

Source: https://huggingface.co/papers/2607.19064 Published on Jul 21

·

Submitted byhttps://huggingface.co/Xinjie-Q

Xinjieon Jul 22

#3 Paper of the day Authors:

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

Abstract

Large-scalevisualgeneratorsareincreasinglycapablebutcostlytotrain,fine-tune,anddeploy.WeintroduceMage-Flow,acompact4B-scalegenerativestackforefficienttext-to-imagegenerationandinstruction-basedimageediting.Thestackisbuiltfromtwoco-designedcomponents:Mage-VAE,alightweighthigh-fidelitylatenttokenizer,andaNative-ResolutionMultimodalDiffusionTransformertrainedwithrectifiedflowmatching.Mage-VAEusesone-stepdiffusion-styleencodinganddecodingwithanchor-latentregularization,preservingthereconstructionqualityofstrongpublicVAEswhilereducingtokenizationcostbymorethananorderofmagnitude.Togetherwithnative-resolutionpackingandstack-levelCUDAkernelfusion,thestacksupportsflexible-resolutiontrainingandimprovesend-to-endtrainingthroughputbyabout2.5times.Builtonthisfoundation,wedevelopacompletemodelfamilywithBase,RL-aligned,andTurbovariantsforbothgenerationandediting.Diffusion-NFTimprovespromptfollowing,textrendering,aestheticquality,andeditingfidelity,whilefew-stepdistillationwithadversarialperceptualguidanceproduces4-stepTurbomodelsforlow-latencyinference.Despiteitscompactscale,Mage-FlowandMage-Flow-Editachievescompetitiveperformanceacrossstandardgenerationandeditingbenchmarks.Moreimportantly,theTurbovariantsmakehigh-resolutiongenerationandeditingpracticalforinteractiveuse:at1024^2resolutiononasingleNVIDIAA100GPU,Mage-Flow-Turbogeneratesanimagein0.59s,andMage-Flow-Edit-Turboeditsanimagein1.02s,whilemaintainingasmallmemoryfootprint.Theseresultsshowthatcarefultokenizer--backbone--systemco-designcandeliverstronghigh-resolutiongenerationandeditingwithinanefficient4Bmodelfamily.

View arXiv pageView PDFProject pageGitHub4Add to collection

Get this paper in your agent:

hf papers read 2607\.19064

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper18

#### microsoft/Mage-Flow Text-to-Image• 4B• Updatedabout 5 hours ago • 22 #### microsoft/Mage-Flow-Base Text-to-Image• 4B• Updatedabout 5 hours ago • 9 #### microsoft/Mage-Flow-Edit-Turbo Image-to-Image• 4B• Updatedabout 5 hours ago • 8 #### microsoft/Mage-Flow-Turbo Text-to-Image• 4B• Updatedabout 5 hours ago • 6 Browse 18 models citing this paper## Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2607.19064 in a dataset README.md to link it from this page.

Spaces citing this paper2

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

microsoft/Mage-Flow

Hugging Face Models Trending

Microsoft releases Mage-Flow, a compact 4B-parameter foundation model for efficient native-resolution text-to-image generation and instruction-based image editing, achieving competitive quality against much larger models.

microsoft/Mage-Flow-Edit-Turbo

Hugging Face Models Trending

Microsoft releases Mage-Flow-Edit-Turbo, a compact 4B-scale generative model for efficient text-to-image generation and instruction-based image editing, achieving state-of-the-art competitive quality through co-designed tokenizer and backbone.

Mage (GitHub Repo)

TLDR AI

Microsoft releases Mage, a family of lightweight 4B-parameter multimodal models for visual understanding and generation, including Mage-VL for image/video understanding and Mage-Flow for text-to-image generation and editing, designed for research and deployment on modest hardware.

Mage-VL: An Efficient Codec-Native Streaming Multimodal Foundation Model

Hugging Face Daily Papers

Mage-VL is an efficient codec-native streaming multimodal foundation model that reduces visual token consumption by over 75% using a custom tokenizer, achieving up to 3.5x inference speedup while matching or outperforming existing models on static and video tasks.