Text-to-Image Generation on GenEval (Compositional Metrics Subset)
98Overall ScoreShow-o
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Show-oCategory=Unified Generative Model2026.04 | 98 | 80 | 66 | 84 | 31 | 50 | 68 | — | |
| SD3.5-M+Flow-GRPO+STARModel Type=Diffusion / Flow Models2026.06 | 97.59 | 100 | 100 | 98 | 96 | 98 | 94 | — | |
| URSA (w/ UDM-GRPO)Model Type=Discrete, #Params=1.7B2026.04 | 96 | 100 | 100 | 95 | 97 | 97 | 85 | — | |
| OctoT2IType=Agentic2026.06 | 96 | 100 | 99 | 95 | 94 | 100 | 86 | — | |
| SD3.5-M+Flow-GRPOModel Category=Flow + RL2026.03 | 95 | 100 | 99 | 95 | 92 | 99 | 86 | — | |
| SD3.5-M (w/ Flow-GRPO)Model Type=Continuous, #Params=2.5B2026.04 | 95 | 100 | 99 | 95 | 92 | 99 | 86 | — | |
| OmniGen2LLM rewriter=false2025.06 | 95 | 99 | 100 | 93 | 91 | 100 | 86 | — | |
| SD3.5-M w/ Flow-GRPOModel Category=Generation-Only2026.05 | 95 | 100 | 99 | 95 | 92 | 99 | 86 | — | |
| SD3.5-M+Flow-GRPOModel Type=Diffusion / Flow Models2026.06 | 95 | 100 | 99 | 95 | 92 | 99 | 86 | — | |
| Ming-Flash-OmniType=Uni.2025.10 | 94 | 99 | 98 | 92 | 94 | 96 | 89 | — | |
| SD3.5-M+CRDModel Category=Flow + RL2026.03 | 93 | 100 | 98 | 92 | 88 | 90 | 73 | — | |
| FUDOKI w/ dFlowGRPOModel Category=Unified Understanding & Generation, Classifier-free guidance (CFG)=false, Denoising steps=82026.05 | 93 | 97 | 93 | 92 | 91 | 88 | 97 | — | |
| Flow-GRPOType=Non-agentic2026.06 | 93 | 99 | 99 | 88 | 95 | 95 | 86 | — | |
| SD3.5-M+DiffusionNFTModel Category=Flow + RL2026.03 | 92 | 100 | 99 | 96 | 82 | 78 | 73 | — | |
| LatentUMVis-Gen + pixel-rewardModel Type=Unified Model, Reward Mechanism=pixel-reward2026.04 | 92 | 99 | 95 | 95 | 92 | 92 | 81 | — | |
| SD3.5-M+AWMModel Category=Flow + RL, CFG usage=w/o CFG2026.03 | 91 | 100 | 98 | 95 | 77 | 78 | 65 | — | |
| Qwen-Image-RLType=Uni.2025.10 | 91 | 100 | 95 | 93 | 92 | 87 | 83 | — | |
| LVRPOType=Unified2026.03 | 91 | 99 | 96 | 87 | 96 | 83 | 81 | — | |
| BLIP3o-NEXT#Params=3B, Steps=302026.04 | 91 | 99 | 95 | 88 | 90 | 92 | 79 | — | |
| BAGEL-RvRType=Refine2026.04 | 91 | 100 | 96 | 91 | 93 | 86 | 80 | — | |
| Lumina-Dimoo w/ selfGRPOModel Category=Unified Understanding & Generation2026.05 | 91 | 100 | 96 | 87 | 95 | 85 | 82 | — | |
| RubricRLBackbone=Qwen-Image (4B)2025.11 | 90.91 | 100 | 96.46 | 85.62 | 97.34 | 82.75 | 83.25 | — | |
| EMF#Params=3B, Steps=42026.04 | 90 | 100 | 94 | 88 | 92 | 91 | 76 | — | |
| TUNAParams.=7B, Model Type=Unified2026.05 | 90 | 100 | 97 | 81 | 91 | 88 | 83 | — | |
| LanceParams.=3B, Model Type=Unified2026.05 | 90 | 100 | 94 | 84 | 97 | 87 | 81 | — | |
| RECAParams=1.5B, trained_with_GPT-4o_data=true2025.09 | 90 | 100 | 97 | 76 | 94 | 91 | 83 | — | |
| RECAParams=1.5B, Training Data=GPT-4o Data2025.09 | 90 | 100 | 97 | 76 | 94 | 91 | 83 | — | |
| Nemotron-Labs-Diffusion-ImageParams=8B, Model Category=Gen. Only2026.06 | 90 | 98 | 93 | 83 | 94 | 88 | 82 | — | |
| RubricRLBackbone=FLUX.1 Dev (12B)2025.11 | 89.69 | 99.69 | 97.47 | 84.38 | 97.07 | 78 | 79.75 | — | |
| LLaDA2.0-UniType=Unified, Arch.=D-Diff. + Diff.2026.04 | 89 | 100 | 98 | 73 | 92 | 90 | 84 | — | |
| MogaoParams.=7B, Model Type=Unified2026.05 | 89 | 100 | 97 | 83 | 93 | 84 | 80 | — | |
| Janus-Pro + GCPOParams=7B2026.05 | 89 | 100 | 95 | 84 | 89 | 83 | 83 | — | |
| Ovis-U1Params=3.6B, trained_with_GPT-4o_data=true2025.09 | 89 | 98 | 98 | 90 | 92 | 79 | 75 | — | |
| Ovis-U1Params=3.6B, Training Data=GPT-4o Data2025.09 | 89 | 98 | 98 | 90 | 92 | 79 | 75 | — | |
| BAGEL#Param=14B, #Data=N/A2026.04 | 88 | 98 | 95 | 84 | 95 | 78 | 77 | — | |
| BAGELLLM rewriter=true2025.06 | 88 | 98 | 95 | 84 | 95 | 78 | 77 | — | |
| Lumina-DiMOOType=Unified, Arch.=D-Diff.2026.04 | 88 | 100 | 94 | 85 | 89 | 85 | 76 | — | |
| Lumina-DimooModel Category=Unified Understanding & Generation2026.05 | 88 | 100 | 94 | 85 | 89 | 85 | 76 | — | |
| BAGELParams.=7B, Model Type=Unified, LLM rewriter=true2026.05 | 88 | 98 | 95 | 84 | 95 | 78 | 77 | — | |
| BAGELParams=14B2026.05 | 88 | 98 | 95 | 84 | 95 | 78 | 77 | — | |
| SeFi-Image-5BParameters=5B2026.06 | 88 | 100 | 92 | 87 | 90 | 84 | 76 | — | |
| IV-CoTExplicit CoT=true, T2I-CompBench Training Split Prompts=false2026.06 | 88 | 100 | 96 | 78 | 92 | 86 | 79 | — | |
| BAGELParams=14B, Model Category=Unified MLLM2026.06 | 88 | 98 | 95 | 84 | 95 | 78 | 77 | — | |
| Lumina-DiMOO + ML-CacheAvg. t (s)=31.952026.03 | 87.83 | 100 | 93.75 | 85.94 | 89.63 | 84.5 | 75.5 | — | |
| Lumina-DiMOO + LADRAvg. t (s)=13.222026.03 | 87.3 | 99.06 | 91.41 | 81.88 | 91.22 | 85.5 | 74.75 | — | |
| Qwen-ImageType=Uni.2025.10 | 87 | 99 | 92 | 89 | 88 | 76 | 77 | — | |
| Stepwise-Flow-GRPOClassifier-Free Guidance (cfg)=4.5, Reward Model=GenEval, Training Compute (GPU hrs)=400 GPU hrs, Training Duration=Extended Training2026.03 | 87 | 99 | 93 | 89 | 87 | 73 | 80 | — | |
| Qwen-ImageModel Type=Specialized Text-to-Image Model2026.04 | 87 | 99 | 92 | 89 | 88 | 76 | 77 | — | |
| LatentUMVis-GenModel Type=Unified Model, Reward Mechanism=self-derived reward2026.04 | 87 | 99 | 92 | 85 | 87 | 89 | 71 | — | |
| Qwen-Image2026.04 | 87 | 99 | 92 | 89 | 88 | 76 | 77 | — | |
| Nucleus-Image2026.04 | 87 | 99 | 95 | 78 | 92 | 85 | 71 | — | |
| Qwen-Image#Param=20B, #Data=N/A2026.04 | 87 | 99 | 92 | 89 | 88 | 76 | 77 | — | |
| TBAC-UniImage-3B#Params=4.6B, Steps=302026.04 | 87 | 99 | 94 | 77 | 92 | 83 | 75 | — | |
| Qwen-Image#Params=20B, Steps=502026.04 | 87 | 99 | 92 | 89 | 88 | 76 | 77 | — | |
| Qwen-ImageType=Gen. Only, Arch.=Diff.2026.04 | 87 | 99 | 92 | 89 | 88 | 76 | 77 | — | |
| LongCat-ImageType=Gen. Only, Arch.=Diff.2026.04 | 87 | 99 | 98 | 86 | 86 | 75 | 73 | — | |
| Qwen-ImageParams.=20B, Model Type=Generation-only2026.05 | 87 | 99 | 92 | 89 | 88 | 76 | 77 | — | |
| TUNA-2Params.=7B, Model Type=Unified2026.05 | 87 | 99 | 96 | 80 | 91 | 84 | 76 | — | |
| Qwen-Image-2507Params=20B2026.05 | 87 | 99 | 92 | 89 | 88 | 76 | 77 | — | |
| Qwen-ImageType=Diff.2026.06 | 87 | — | 92 | — | — | 76 | 77 | — | |
| SeFi-Image-2BParameters=2B2026.06 | 87 | 99 | 93 | 84 | 91 | 78 | 78 | — | |
| SeFi-Image-1BParameters=1B2026.06 | 87 | 99 | 91 | 83 | 92 | 82 | 75 | — | |
| Qwen-Image2026.06 | 87 | 99 | 92 | 89 | 88 | 76 | 77 | — | |
| SeFi-Image-5BModel Parameters=5B, Turbo Variant=false2026.06 | 87 | 100 | 92 | 85 | 90 | 81 | 77 | — | |
| SeFi-Image-2BModel Parameters=2B, Turbo Variant=false2026.06 | 87 | 99 | 93 | 84 | 91 | 78 | 78 | — | |
| SeFi-Image-1BModel Parameters=1B, Turbo Variant=false2026.06 | 87 | 99 | 91 | 83 | 92 | 82 | 75 | — | |
| TUNA-2Explicit CoT=false, T2I-CompBench Training Split Prompts=false2026.06 | 87 | 99 | 96 | 80 | 91 | 84 | 76 | — | |
| Qwen-Image-2507Params=20B, Model Category=Gen. Only2026.06 | 87 | 99 | 92 | 89 | 88 | 76 | 77 | — | |
| Lumina-DiMOOAvg. t (s)=57.012026.03 | 86.66 | 97.5 | 93.94 | 85 | 91.49 | 79 | 73 | — | |
| AD-LoopCapability=Und. and Gen.2026.02 | 86 | 98 | 94 | 83 | 90 | 80 | 74 | — | |
| OmniGen2Type=Uni.2025.10 | 86 | 99 | 96 | 74 | 98 | 71 | 75 | — | |
| UAE2025.11 | 86 | — | — | — | — | — | — | — | |
| RecA2025.11 | 86 | 100 | 98 | 71 | 93 | 76 | 77 | — | |
| DeCo-XXL/16Diffusion Params=1.1B, Resolution=512×5122025.11 | 86 | 100 | 92 | 72 | 91 | 80 | 79 | — | |
| BitDance#Param=14B, #Data=N/A2026.04 | 86 | 100 | 96 | 71 | 95 | 72 | 83 | — | |
| FiMRstep=3rd2026.04 | 86 | 99 | 95 | 70 | 94 | 84 | 76 | — | |
| BLIP3o-NEXT#Params=3B, Steps=42026.04 | 86 | 99 | 93 | 84 | 84 | 86 | 70 | — | |
| Emu3.5Model Type=Discrete, #Params=34B2026.04 | 86 | — | — | — | — | — | — | — | |
| LLaDA-oType=Unified, Arch.=D-Diff. + Diff.2026.04 | 86 | 99 | 98 | 73 | 96 | 69 | 83 | — | |
| Janus-Pro-R1Params=7B2026.05 | 86 | 99 | 94 | 66 | 92 | 87 | 78 | — | |
| ARM-RLType=AR2026.06 | 86 | — | 93 | — | — | 89 | 90 | — | |
| SeFi-Image-5B-TurboModel Parameters=5B, Turbo Variant=true2026.06 | 86 | 99 | 93 | 81 | 90 | 84 | 71 | — | |
| OpenUni-L-1024Explicit CoT=false, T2I-CompBench Training Split Prompts=false2026.06 | 86 | 99 | 92 | 76 | 91 | 82 | 77 | — | |
| DracoExplicit CoT=true, T2I-CompBench Training Split Prompts=false2026.06 | 86 | 100 | 99 | 81 | 91 | 70 | 76 | — | |
| RECAParams=1.5B2025.09 | 86 | 100 | 98 | 71 | 93 | 76 | 77 | — | |
| RECAParams=1.5B2025.09 | 86 | 100 | 98 | 71 | 93 | 76 | 77 | — | |
| X-OmniBackbone=Qwen-Image (4B)2025.11 | 85.35 | 99.38 | 93.69 | 74.38 | 94.41 | 77.25 | 73 | — | |
| CLIPScoreBackbone=Qwen-Image (4B)2025.11 | 85.21 | 99.38 | 93.94 | 72.19 | 96.28 | 76.25 | 73.25 | — | |
| CLIPScoreBackbone=FLUX.1 Dev (12B)2025.11 | 85.18 | 98.75 | 96.97 | 80.62 | 95.21 | 72.5 | 67 | — | |
| AR-GRPOBackbone=Qwen-Image (4B)2025.11 | 85.16 | 98.44 | 93.69 | 77.19 | 94.41 | 74.5 | 72.75 | — | |
| LatentUMBaseModel Type=Unified Model2026.04 | 85 | 99 | 92 | 72 | 91 | 83 | 73 | — | |
| CAFMPE=Yes, CFG=Yes, Resolution=512px2026.04 | 85 | 99 | 92 | 81 | 94 | 71 | 71 | — | |
| FiMRstep=2nd2026.04 | 85 | 99 | 95 | 65 | 94 | 83 | 76 | — | |
| EMF#Params=3B, Steps=22026.04 | 85 | 99 | 91 | 81 | 86 | 86 | 66 | — | |
| InternVL-UType=Unified, Arch.=AR + Diff.2026.04 | 85 | 99 | 94 | 74 | 91 | 77 | 74 | — | |
| BAGEL-RvE (IRG)Type=Refine2026.04 | 85 | 98 | 94 | 83 | 86 | 74 | 73 | — | |
| InternVL-UParams.=1.7B, Model Type=Unified2026.05 | 85 | 99 | 94 | 74 | 91 | 77 | 74 | — | |
| Janus-Pro + GRPOParams=7B2026.05 | 85 | 99 | 93 | 81 | 83 | 83 | 73 | — | |
| FLUX.2-Klein-9BParameters=9B2026.06 | 85 | 100 | 96 | 85 | 91 | 70 | 68 | — | |
| ILLUME-X# Params=7+7 B2026.06 | 85 | 99 | 95 | 81 | 92 | 71 | 74 | — |