Text-to-image generation on GenEval (Detailed Compositional Metrics)
96Overall ScoreLatentMorph
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| LatentMorphBackbone=Janus-Pro2026.02 | 96 | — | 97 | — | — | 98 | 92 | — | |
| BAGEL + R32026.02 | 96 | 100 | 100 | 95 | 95 | 98 | 89 | — | |
| UniDFlowParams=4B2026.02 | 95 | — | — | — | — | — | — | — | |
| MILRBackbone=Janus-Pro2026.02 | 95 | — | 96 | — | — | 98 | 91 | — | |
| FlowGRPO2026.02 | 95 | 100 | 99 | 95 | 92 | 99 | 86 | — | |
| SIDiffAgentepoch=22026.02 | 94 | 99 | 98 | 93 | 100 | 81 | 94 | — | |
| EMMAParams=4B2026.02 | 93 | — | — | — | — | — | — | — | |
| EMMAprompt rewriting=true2025.12 | 93 | 100 | 99 | 87 | 98 | 86 | 87 | — | |
| Qwen-Image(-RL)Params=7B+20B2026.02 | 91 | — | — | — | — | — | — | — | |
| STAR-7BModel Parameters=7B2025.12 | 91 | 98 | 94 | 90 | 92 | 91 | 80 | — | |
| Qwen-Image-RLRL=true2025.12 | 91 | 100 | 95 | 93 | 92 | 87 | 83 | — | |
| EMMAprompt rewriting=false2025.12 | 91 | 100 | 98 | 83 | 96 | 83 | 85 | — | |
| Qwen-Image-RL2026.02 | 91 | 100 | 95 | 93 | 92 | 87 | 83 | — | |
| Lumina-DiMOO# Params=8B, Model Type=Unified, Prompting Protocol=rewritten prompts2026.03 | 91 | 100 | 96 | 87 | 95 | 85 | 82 | — | |
| SPAR-3BParams=3B2026.06 | 91 | — | — | 84 | — | 87 | — | — | |
| MUDDITParams=8B2026.02 | 90 | — | — | — | — | — | — | — | |
| TUNADate=2025.12, Parameters=7B, Tokenization=Continuous2026.02 | 90 | 100 | 97 | 81 | 91 | 88 | 83 | — | |
| UniReasonReasoning=w Reasoning2026.02 | 90 | 100 | 96 | 82 | 90 | 88 | 82 | — | |
| Lumina-DiMOO (Ours)Generation step=642026.02 | 90 | 99 | 94 | 88 | 93 | 87 | 79 | — | |
| SIDiffAgent2026.02 | 89 | 99 | 97 | 80 | 99 | 71 | 92 | — | |
| MogaoReasoning=w/o Reasoning2026.02 | 89 | 100 | 97 | 83 | 93 | 84 | 80 | — | |
| ThinkGen*Model Category=Und. and Gen., CoT reasoning=true2025.12 | 89 | — | — | 84 | — | 80 | — | — | |
| MogaoModel Type=Unified Models2025.12 | 89 | 100 | 97 | 83 | 93 | 84 | 80 | — | |
| Ovis-U1Model Type=Unified Models2025.12 | 89 | 98 | 98 | 90 | 92 | 79 | 75 | — | |
| Mogao2025.12 | 89 | 100 | 97 | 83 | 93 | 84 | 80 | — | |
| Mogao# Params=7B, Model Type=Unified, Prompting Protocol=rewritten prompts2026.03 | 89 | 100 | 97 | 83 | 93 | 84 | 80 | — | |
| LaViDa-OModel Type=Unified, Prompting Protocol=rewritten prompts2026.03 | 89 | — | — | — | — | — | — | — | |
| SPAR-1BParams=1B2026.06 | 89 | — | — | 83 | — | 85 | — | — | |
| SeGroSUMM=Harmon-1.5B, Fine-Tuning Dataset=BLIP3o-60k Dataset2026.03 | 88.66 | 100 | 97.7 | 79.1 | 90.4 | 83.5 | 81.3 | — | |
| BAGELType=Unified, LLM rewriter=true2025.05 | 88 | 98 | 95 | 84 | 95 | 78 | 77 | — | |
| Qwen-Agents2026.02 | 88 | 96 | 95 | 84 | 96 | 70 | 89 | — | |
| BagelParams=7B+7B2026.02 | 88 | — | — | — | — | — | — | — | |
| BAGELType=Unified, LLM rewriter=true2026.02 | 88 | 98 | 95 | 84 | 95 | 78 | 77 | — | |
| Lumina-DiMOOReasoning=w/o Reasoning2026.02 | 88 | 100 | 94 | 85 | 89 | 85 | 76 | — | |
| BAGELReasoning=w Reasoning2026.02 | 88 | 98 | 95 | 84 | 95 | 78 | 77 | — | |
| BAGELType=Unified, LLM Rewriter=true2026.01 | 88 | 98 | 95 | 84 | 95 | 78 | 77 | — | |
| BAGELArchitecture=AR + Diffusion Models, LLM rewriter=true2026.01 | 88 | 98 | 95 | 84 | 95 | 78 | 77 | — | |
| ThinkGenModel Category=Und. and Gen., CoT reasoning=false2025.12 | 88 | — | — | 81 | — | 79 | — | — | |
| BAGELModel Type=Unified Models2025.12 | 88 | 98 | 95 | 84 | 95 | 78 | 77 | — | |
| BAGELprompt rewriting=true2025.12 | 88 | 98 | 95 | 84 | 95 | 78 | 77 | — | |
| BAGEL# Params=7B MoT, Model Type=Unified, Prompting Protocol=rewritten prompts2026.03 | 88 | 98 | 95 | 84 | 95 | 78 | 77 | — | |
| KelixParameters=8B, Tokenization=Discrete, Resolution=1024 x 10242026.02 | 87.6 | 100 | 93 | 81 | 87 | 85 | 79 | — | |
| Qwen-Image2026.02 | 87 | 96 | 92 | 83 | 88 | 71 | 91 | — | |
| MammothModa-2Params=8B+3B+2B2026.02 | 87 | — | — | — | — | — | — | — | |
| Qwen-ImageDate=2025.08, Parameters=27B, Category=LLM & Diffusion Conjunction2026.02 | 87 | 99 | 92 | 89 | 88 | 76 | 77 | — | |
| Qwen-ImageReasoning=w/o Reasoning2026.02 | 87 | 99 | 92 | 89 | 88 | 76 | 77 | — | |
| Qwen-ImageArchitecture=Diffusion Models, LLM rewriter=false2026.01 | 87 | 99 | 92 | 89 | 88 | 76 | 77 | — | |
| Qwen-ImageRL=false2025.12 | 87 | 99 | 92 | 89 | 88 | 76 | 77 | — | |
| Lumina-DiMOO (ReMDM)Generation step=642026.02 | 87 | 100 | 94 | 86 | 87 | 82 | 74 | — | |
| TwiG-ZSBackbone=Janus-Pro, Mode=Zero-shot2026.02 | 86 | — | 94 | — | — | 84 | 67 | — | |
| CoF-T2IModel Category=Video Models2026.01 | 86 | 98 | 95 | 83 | 89 | 83 | 71 | — | |
| EMU3.5Architecture=AR Models, LLM rewriter=true2026.01 | 86 | — | — | — | — | — | — | — | |
| OmniGen2Model Type=Gen. Only Models2025.12 | 86 | 99 | 96 | 74 | 98 | 72 | 75 | — | |
| STAR-3BModel Parameters=3B2025.12 | 86 | 98 | 87 | 85 | 91 | 79 | 76 | — | |
| OminiGen2prompt rewriting=true2025.12 | 86 | 99 | 96 | 74 | 98 | 71 | 75 | — | |
| Lumina-DiMOOGeneration step=64, Type=Pre-trained baseline2026.02 | 86 | 99 | 93 | 85 | 84 | 84 | 71 | — | |
| LLaDA-o# Params=8B MoT, Model Type=Unified, Prompting Protocol=rewritten prompts2026.03 | 86 | 99 | 98 | 73 | 96 | 69 | 83 | — | |
| RecaUMM=Harmon-1.5B, Fine-Tuning Dataset=BLIP3o-60k Dataset2026.03 | 85.2 | — | — | — | — | — | — | — | |
| Ming-UnivisionDate=2025.10, Parameters=16B, Tokenization=Continuous2026.02 | 85 | 100 | 93 | 59 | 93 | 92 | 70 | — | |
| ManzanoDate=2025.09, Parameters=30B, Tokenization=Hybrid2026.02 | 85 | 100 | 91 | 83 | 87 | 84 | 65 | — | |
| ChatUMMType=Unified, LLM rewriter=true2026.02 | 85 | 99 | 96 | 79 | 92 | 69 | 77 | — | |
| IRGReasoning=w Reasoning2026.02 | 85 | 98 | 94 | 83 | 86 | 74 | 73 | — | |
| T2I-CopilotBackbone=Janus-Pro2026.02 | 85 | — | 94 | — | — | 86 | 66 | — | |
| SFT*UMM=Harmon-1.5B, Fine-Tuning Dataset=BLIP3o-60k Dataset2026.03 | 85 | 99.1 | 94.7 | 77.8 | 86.4 | 78.3 | 74 | — | |
| OpenUni-LParams=3B2026.06 | 85 | — | — | 77 | — | 75 | — | — | |
| GPT Image 1mode=High2026.02 | 84 | 99 | 92 | 85 | 92 | 75 | 61 | — | |
| Seedream 3.02026.02 | 84 | 99 | 96 | 91 | 93 | 47 | 80 | — | |
| GPT-4oDate=2024.05, Category=Proprietary Models2026.02 | 84 | 99 | 92 | 85 | 92 | 75 | 61 | — | |
| GPT-4o2026.02 | 84 | 99 | 92 | 85 | 92 | 75 | 61 | — | |
| Seedream 4.02026.02 | 84 | 99 | 92 | 72 | 91 | 76 | 74 | — | |
| BLIP3-oReasoning=w/o Reasoning2026.02 | 84 | — | — | — | — | — | — | — | |
| TIRBackbone=Janus-Pro2026.02 | 84 | — | 93 | — | — | 84 | 68 | — | |
| BLIP3-o 8BModel Category=Unified MLLMs2026.01 | 84 | — | — | — | — | — | — | — | |
| GPT Image 1 [High]Architecture=Diffusion Models, LLM rewriter=false2026.01 | 84 | 99 | 92 | 85 | 92 | 75 | 61 | — | |
| Seedream 3.0Architecture=Diffusion Models, LLM rewriter=false2026.01 | 84 | 99 | 96 | 91 | 93 | 47 | 80 | — | |
| NextFlow-RLArchitecture=AR Models, LLM rewriter=true2026.01 | 84 | 100 | 92 | 75 | 90 | 76 | 70 | — | |
| BLIP3-o-8BModel Category=Und. and Gen.2025.12 | 84 | — | — | — | — | — | — | — | |
| UniWorld-V1Model Type=Unified Models2025.12 | 84 | 99 | 93 | 81 | 89 | 74 | 71 | — | |
| GPT-4oModel Type=Unified Models2025.12 | 84 | 99 | 92 | 85 | 92 | 75 | 61 | — | |
| BLIP3-oprompt rewriting=true2025.12 | 84 | — | — | — | — | — | — | — | |
| BLIP3-o-8BParams=7B2026.06 | 84 | — | — | — | — | — | — | — | |
| OpenUni-BParams=1B2026.06 | 84 | — | — | 74 | — | 77 | — | — | |
| TarParams=7B2026.06 | 84 | — | — | 83 | — | 80 | — | — | |
| BLIP3-o 8BType=Unified Model, LLM rewriter=true2025.10 | 84 | — | — | — | — | — | — | — | |
| HiDream-I1-Full2026.02 | 83 | 100 | 98 | 79 | 91 | 60 | 72 | — | |
| X-OmniDate=2025.07, Parameters=7B, Tokenization=Discrete2026.02 | 83 | 98 | 95 | 75 | 91 | 71 | 68 | — | |
| NextflowDate=2026.01, Parameters=7B, Tokenization=Discrete2026.02 | 83 | 98 | 92 | 73 | 90 | 77 | 69 | — | |
| Mind-OmniReasoning=w Reasoning2026.02 | 83 | 99 | 94 | 71 | 90 | 71 | 71 | — | |
| Uni-CoTReasoning=w Reasoning2026.02 | 83 | 99 | 96 | 84 | 92 | 57 | 71 | — | |
| Self-CoTBackbone=Janus-Pro2026.02 | 83 | — | 90 | — | — | 78 | 70 | — | |
| LASAGNAType=Unified, LLM Rewriter=true2026.01 | 83 | 99 | 97 | 78 | 83 | 74 | 65 | — | |
| Multimodal RWRModality Ablation=Multimodal, Reward Configuration=Joint Multimodal2026.01 | 83 | 99 | 97 | 82 | 88 | 61 | 72 | — | |
| HiDream-I1-FullArchitecture=Diffusion Models, LLM rewriter=false2026.01 | 83 | 100 | 98 | 79 | 91 | 60 | 72 | — | |
| NextFlowArchitecture=AR Models, LLM rewriter=true2026.01 | 83 | 98 | 92 | 73 | 90 | 77 | 69 | — | |
| X-OmniModel Type=Unified Models2025.12 | 83 | 98 | 95 | 75 | 91 | 71 | 68 | — | |
| HiDream-I1-Full2025.12 | 83 | 100 | 98 | 79 | 91 | 60 | 72 | — | |
| FLUX.1-devType=Gen. Only2025.05 | 82 | 98 | 93 | 75 | 93 | 68 | 65 | — | |
| BAGELType=Unified, LLM rewriter=false2025.05 | 82 | 99 | 94 | 81 | 88 | 64 | 63 | — | |
| FLUX.1-devDate=2024.08, Parameters=12B, Category=Dedicated T2I Models2026.02 | 82 | 98 | 93 | 75 | 93 | 68 | 65 | — | |
| BagelDate=2025.05, Parameters=14B, Tokenization=Continuous2026.02 | 82 | 99 | 94 | 81 | 88 | 64 | 63 | — |