← Unified Models

Model Architectures

The full spectrum of unified multimodal model designs — diffusion-based, autoregressive, hybrid, and any-to-any architectures.

⌘K

Approaches are categorized by their backbone architecture — following the taxonomy from Awesome-Unified-Multimodal-Models. Includes diffusion-based, autoregressive (MLLM), hybrid AR-diffusion, and any-to-any architectures.

Diffusion-Based Unified Models 7 papers

ModelFull TitleVenueDateCode
UniModelA Visual-Only Framework for Unified Multimodal Understanding and Generation arXiv2025/11—
Lavida-OElastic Large Masked Diffusion Models for Unified Multimodal Understanding and Generation arXiv2025/09✅
MudditLiberating Generation Beyond Text-to-Image with a Unified Discrete Diffusion Model arXiv2025/05✅
FUDOKIDiscrete Flow-based Unified Understanding and Generation via Kinetic-Optimal Velocities arXiv2025/05—
MMaDAMultimodal Large Diffusion Language Models arXiv2025/05✅
UniDiscUnified Multimodal Discrete Diffusion arXiv2025/03✅
Dual DiffusionDual Diffusion for Unified Image Generation and Understanding arXiv2024/12✅

MLLM Autoregressive — Pixel Encoding 14 papers

ModelFull TitleVenueDateCode
Emu3.5Native Multimodal Models are World Learners arXiv2025/10✅
OneCatDecoder-Only Auto-Regressive Model for Unified Understanding and Generation arXiv2025/09✅
SelftokDiscrete Visual Tokens of Autoregression, by Diffusion, and for Reasoning arXiv2025/05✅
TokLIPMarry Visual Tokens to CLIP for Multimodal Comprehension and Generation arXiv2025/05✅
HarmonHarmonizing Visual Representations for Unified Multimodal Understanding and Generation arXiv2025/03✅
UGenUnified Autoregressive Multimodal Model with Progressive Vocabulary Learning arXiv2025/03—
SynerGen-VLTowards Synergistic Image Understanding and Generation with Vision Experts and Token Folding arXiv2024/12—
LiquidLanguage Models are Scalable and Unified Multi-modal Generators arXiv2024/12✅
OrthusAutoregressive Interleaved Image-Text Generation with Modality-Specific Heads arXiv2024/11✅
ANOLEAn Open, Autoregressive, Native Large Multimodal Model for Interleaved Image-Text Generation arXiv2024/07✅
ChameleonMixed-Modal Early-Fusion Foundation Models arXiv2024/05✅
LWMWorld Model on Million-Length Video And Language With Blockwise RingAttention ICLR2024/02✅
Emu3Emu3: Next-Token Prediction is All You Need arXiv2024/09✅
EmuGenerative Pretraining in Multimodality ICLR2023/07✅

MLLM Autoregressive — Semantic Encoding 22 papers

ModelFull TitleVenueDateCode
UniCornTowards Self-Improving Unified Multimodal Models through Self-Generated Supervision arXiv2026/01—
ORIONORION: Decoupling and Alignment for Unified Autoregressive Understanding and Generation ICLR2026—
Ming-UniVisionJoint Image Understanding and Generation with a Unified Continuous Tokenizer arXiv2025/10✅
Bifrost-1Bridging Multimodal LLMs and Diffusion Models with Patch-level CLIP Latents arXiv2025/08✅
Qwen-ImageQwen-Image Technical Report arXiv2025/08✅
X-OmniReinforcement Learning Makes Discrete Autoregressive Image Generative Models Great Again arXiv2025/07✅
Ovis-U1Ovis-U1 Technical Report arXiv2025/06✅
UniCode²Cascaded Large-scale Codebooks for Unified Multimodal Understanding and Generation arXiv2025/06—
OmniGen2Exploration to Advanced Multimodal Generation arXiv2025/06✅
TarVision as a Dialect: Unifying Visual Understanding and Generation via Text-Aligned Representations arXiv2025/06✅
UniForkExploring Modality Alignment for Unified Multimodal Understanding and Generation arXiv2025/06✅
UniWorldHigh-Resolution Semantic Encoders for Unified Visual Understanding and Generation arXiv2025/06✅
PiscesAn Auto-regressive Foundation Model for Image Understanding and Generation arXiv2025/06✅
DualTokenTowards Unifying Visual Understanding and Generation with Dual Visual Vocabularies arXiv2025/03✅
UniTokA Unified Tokenizer for Visual Generation and Understanding arXiv2025/02✅
QLIPText-Aligned Visual Tokenization Unifies AR Multimodal Understanding and Generation arXiv2025/02✅
MetaMorphMultimodal Understanding and Generation via Instruction Tuning arXiv2024/12✅
ILLUMEIlluminating Your LLMs to See, Draw, and Self-Enhance arXiv2024/12—
PUMAEmpowering Unified MLLM with Multi-granular Visual Generation arXiv2024/10✅
VILA-UUnified Foundation Model Integrating Visual Understanding and Generation ICLR2024/09✅
Mini-GeminiMining the Potential of Multi-modality Vision Language Models arXiv2024/03✅
Emu2Generative Multimodal Models are In-Context Learners CVPR2023/12✅

MLLM Autoregressive — Learnable Query Encoding 11 papers

ModelFull TitleVenueDateCode
UniPic 2.0Building Kontext Model with Online RL for Unified Multimodal Model arXiv2025/09✅
TBAC-UniImageUnified Understanding and Generation by Ladder-Side Diffusion Tuning arXiv2025/08✅
UniLIPAdapting CLIP for Unified Multimodal Understanding, Generation and Editing arXiv2025/07—
OpenUniA Simple Baseline for Unified Multimodal Understanding and Generation arXiv2025/05✅
BLIP3-oA Family of Fully Open Unified Multimodal Models arXiv2025/05✅
Ming-Lite-UniAdvancements in Unified Architecture for Natural Multimodal Interaction arXiv2025/05✅
Nexus-GenA Unified Model for Image Understanding, Generation, and Editing arXiv2025/04✅
MetaQueriesTransfer between Modalities with MetaQueries arXiv2025/04—
SEED-XMultimodal Models with Unified Multi-granularity Comprehension and Generation arXiv2024/04✅
SEED-LLaMAMaking LLaMA SEE and Draw with SEED Tokenizer ICLR2023/10✅
SEEDPlanting a SEED of Vision in Large Language Model arXiv2023/07✅

Hybrid Encoding — Pseudo & Joint 14 papers

ModelFull TitleArchitectureVenueDate
Skywork UniPicUnified Autoregressive Modeling for Visual Understanding and Generation PseudoarXiv2025/08
MindOmniUnleashing Reasoning Generation in Vision Language Models with RGPO PseudoarXiv2025/05
UniFluidUnified Autoregressive Visual Generation and Understanding with Continuous Tokens PseudoarXiv2025/03
OmniMambaEfficient and Unified Multimodal Understanding and Generation via State Space Models PseudoarXiv2025/03
Janus-ProUnified Multimodal Understanding and Generation with Data and Model Scaling PseudoarXiv2025/01
JanusDecoupling Visual Encoding for Unified Multimodal Understanding and Generation PseudoarXiv2024/10
Show-o2Improved Native Unified Multimodal Models JointarXiv2025/06
UniTokenHarmonizing Multimodal Understanding and Generation through Unified Visual Encoding JointCVPRW2025/04
VARGPT-v1.1Improve Visual AR Large Unified Model via Iterative Instruction Tuning and RL JointarXiv2025/04
ILLUME+Illuminating Unified MLLM with Dual Visual Tokenization and Diffusion Refinement JointarXiv2025/04
VARGPTUnified Understanding and Generation in a Visual Autoregressive MLLM JointarXiv2025/01
TokenFlowUnified Image Tokenizer for Multimodal Understanding and Generation JointCVPR2024/12
MUSE-VLModeling Unified VLM through Semantic Discrete Encoding JointarXiv2024/11
SemHiTokSemantic-Guided Hierarchical Codebook for Multimodal Understanding and Generation JointarXiv2025/03

MLLM AR-Diffusion Architectures 13 papers

ModelFull TitleVenueDateCode
MammothModa2A Unified AR-Diffusion Framework for Multimodal Understanding and Generation arXiv2025/11✅
EMMAEfficient Multimodal Understanding, Generation, and Editing with a Unified Architecture arXiv2025/12✅
HBridgeH-Shape Bridging of Heterogeneous Experts for Unified Multimodal Understanding and Generation arXiv2025/11—
TUNATaming Unified Visual Representations for Native Unified Multimodal Models arXiv2025/12✅
LightFusionA Light-weighted, Double Fusion Framework for Unified Multimodal Understanding and Generation arXiv2025/10✅
BAGELEmerging Properties in Unified Multimodal Pretraining arXiv2025/05✅
MogaoAn Omni Foundation Model for Interleaved Multi-Modal Generation arXiv2025/05—
LMFusionAdapting Pretrained Language Models for Multimodal Generation arXiv2024/12—
MonoFormerOne Transformer for Both Diffusion and Autoregression arXiv2024/09✅
Show-oOne Single Transformer to Unify Multimodal Understanding and Generation ICLR2024/08✅
TransfusionPredict the Next Token and Diffuse Images with One Multi-Modal Model ICLR2024/08✅
JanusFlowHarmonizing Autoregression and Rectified Flow for Unified Multimodal Understanding and Generation arXiv2024/11✅
UniRealUniversal Image Generation and Editing via Learning Real-World Dynamics arXiv2025—

Any-to-Any Multimodal Models 14 papers

ModelFull TitleVenueDateCode
LongCat-Flash-OmniLongCat-Flash-Omni Technical Report arXiv2025/11✅
Ming-Flash-OmniA Sparse, Unified Architecture for Multimodal Perception and Generation arXiv2025/10✅
Qwen3-OmniQwen3-Omni Technical Report arXiv2025/09✅
Ming-OmniA Unified Multimodal Model for Perception and Generation arXiv2025/06✅
M2-omniAdvancing Omni-MLLM for Comprehensive Modality Support arXiv2025/02—
OmniFlowAny-to-Any Generation with Multi-Modal Rectified Flows CVPR2024/12✅
SpiderAny-to-Many Multimodal LLM arXiv2024/11✅
MIOA Foundation Model on Multimodal Tokens arXiv2024/09✅
X-VILACross-Modality Alignment for Large Language Model arXiv2024/05—
AnyGPTUnified Multimodal LLM with Discrete Sequence Modeling arXiv2024/02✅
Video-LaVITUnified Video-Language Pre-training with Decoupled Visual-Motional Tokenization ICML2024/02✅
Unified-IO 2Scaling Autoregressive Multimodal Models with Vision, Language, Audio, and Action CVPR2023/12✅
NExT-GPTAny-to-Any Multimodal LLM ICML2023/09✅
ImageBindOne Embedding Space To Bind Them All CVPR2023✅

Applications & Opportunities 4 papers

ModelFull TitleVenueDateCode
UniGameTurning a Unified Multimodal Model Into Its Own Adversary arXiv2025/11✅
UniCTokensBoosting Personalized Understanding and Generation via Unified Concept Tokens arXiv2025/05✅
T2I-R1Reinforcing Image Generation with Collaborative Semantic-level and Token-level CoT arXiv2025/01✅
Fair-UMLLMOn Fairness of Unified Multimodal Large Language Model for Image Generation arXiv2025/02—