Text-to-Image Generation on DPG
95.7Overall ScoreSIDiffAgent
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| SIDiffAgentepoch=ep22026.02 | 95.7 | 93.92 | 97.93 | 97.23 | 93.97 | 95.2 | — | |
| SIDiffAgent2026.02 | 93.68 | 90.58 | 96.77 | 96.11 | 92.07 | 93.6 | — | |
| HiDream-O1-Image-ProNumber of Parameters=200B+2026.05 | 90.3 | 94.97 | 95.42 | 92.59 | 90.82 | 89.5 | — | |
| HiDream-O1-ImageNumber of Parameters=8B2026.05 | 89.83 | 95.15 | 92.32 | 93.74 | 92.88 | 90.25 | — | |
| Amber-Image-10BParameters=10B2026.02 | 89.61 | 83.28 | 92.54 | 90.16 | 94.47 | 87.6 | — | |
| Qwen-Agents2026.02 | 89.51 | 90.58 | 94.54 | 92.93 | 89.48 | 86.4 | — | |
| Prompt ReinjectionBackbone=Qwen-Image2026.02 | 89.33 | 85.11 | 94.12 | 91.09 | 95.2 | 88 | — | |
| Qwen ImageBackbone=Qwen-Image2026.02 | 89.2 | 84.8 | 93.98 | 91.09 | 95.08 | 89.6 | — | |
| Amber-Image-6BParameters=6B2026.02 | 88.96 | 79.73 | 90.45 | 91.64 | 93.87 | 89.11 | — | |
| Base modelModel Context=Qwen-Image, P.(B)↓=20, M.(%)↓=100, L.(ms)↓=26252025.11 | 88.9 | — | — | — | — | — | 0 | |
| Seedream-4.02026.05 | 88.63 | 87.17 | 92.41 | 92.29 | 93.33 | 95.48 | — | |
| Qwen-ImageReasoning capability=w/o Reasoning2026.02 | 88.32 | 91.32 | 91.56 | 92.02 | 94.31 | 92.73 | — | |
| Qwen-ImageModel Type=Diffusion Model2026.01 | 88.32 | 91.32 | 91.56 | 92.02 | 94.31 | 92.73 | — | |
| NextFlow-RLModel Type=AR Model, Fine-tuning=RL-finetuned2026.01 | 88.32 | 92.09 | 93.48 | 91.92 | 92.08 | 93.22 | — | |
| Qwen-Image2026.02 | 88.32 | 91.32 | 91.56 | 92.02 | 94.31 | 92.73 | — | |
| Qwen-ImageType=Gen. Only, Arch.=Diff.2026.04 | 88.32 | 91.32 | 91.56 | 92.02 | 94.31 | 92.73 | — | |
| Qwen-ImageNumber of Parameters=7B + 20B2026.05 | 88.32 | 91.32 | 91.56 | 92.02 | 94.31 | 92.73 | — | |
| Qwen-ImageType=Diff.2026.06 | 88.32 | 91.32 | — | — | 94.31 | — | — | |
| Seedream 3.02026.02 | 88.27 | 94.31 | 92.65 | 91.36 | 92.78 | 88.24 | — | |
| Seedream 3.0Model Type=Diffusion Model, Version=3.02026.01 | 88.27 | 94.31 | 92.65 | 91.36 | 92.78 | 88.24 | — | |
| Seedream 3.02026.02 | 88.27 | 94.31 | 92.65 | 91.36 | 92.78 | 88.24 | — | |
| Seedream 3.0Type=Gen. Only, Arch.=Diff.2026.04 | 88.27 | 94.31 | 92.65 | 91.36 | 92.78 | 88.24 | — | |
| EMU3.5Model Type=AR Model, Version=3.52026.01 | 88.26 | — | — | — | — | — | — | |
| Seedream 4.02026.02 | 88.25 | 94.1 | 92.28 | 92.75 | 93.67 | 92.77 | — | |
| Z-Image2026.02 | 88.14 | 93.39 | 91.22 | 93.16 | 92.22 | 91.52 | — | |
| Z-ImageParam.=6B, Resolution=1024 x 10242026.03 | 88.1 | 93.4 | 91.2 | 93.2 | 92.2 | 91.5 | — | |
| PPCL(14B)Model Context=Qwen-Image, P.(B)↓=14, M.(%)↓=79.4, L.(ms)↓=17922025.11 | 87.9 | — | — | — | — | — | 0.42 | |
| Qwen-Image2026.02 | 87.84 | 91.19 | 92.58 | 92.61 | 87.54 | 85.2 | — | |
| LLaDA2.0-UniType=Unified, Arch.=D-Diff. + Diff.2026.04 | 87.76 | 91.14 | 93.55 | 91.98 | 92.17 | 93.18 | — | |
| FLUX.2 [Dev]Number of Parameters=24B + 32B2026.05 | 87.57 | 92.2 | 91.36 | 93.28 | 93.52 | 89.72 | — | |
| Lumina-Image 2.02026.02 | 87.2 | — | 91.97 | 90.2 | 94.85 | — | — | |
| Lumina-Image 2.0Type=Gen. Only, Arch.=Diff.2026.04 | 87.2 | 86.63 | 91.97 | 90.2 | 94.85 | 84.8 | — | |
| Lumina-Image 2.0Type=Diff.2026.06 | 87.2 | — | — | — | 94.85 | — | — | |
| Prompt ReinjectionBackbone=SD3-medium2026.02 | 87.12 | 86.34 | 92.28 | 89.44 | 94.39 | 87.2 | — | |
| LLaDA-oType=Unified, Arch.=D-Diff. + Diff.2026.04 | 87.04 | 92.91 | 93.3 | 90.4 | 91.75 | 92.79 | — | |
| Prompt ReinjectionBackbone=SD3.5-large2026.02 | 86.99 | 83.89 | 91.9 | 90.09 | 94.31 | 85.6 | — | |
| Nano Banana 2.02026.05 | 86.9 | 85.17 | 92.55 | 91.16 | 90.45 | 91.08 | — | |
| UniCornModel Category=Unified Multimodal Models2026.01 | 86.8 | — | — | — | — | — | — | |
| LongCat-Image2026.02 | 86.8 | 89.1 | 92.54 | 92 | 93.28 | 87.5 | — | |
| LongCat-ImageType=Gen. Only, Arch.=Diff.2026.04 | 86.8 | 89.1 | 92.54 | 92 | 93.28 | 87.5 | — | |
| CAR (8B)Type=Unidirectional (NCP*), Parameters=8B, Prompt rewriting / self-CoT=false2026.05 | 86.72 | 89.4 | — | — | 94.16 | — | — | |
| PPCL(Fine-tuning)Model Context=Qwen-Image, P.(B)↓=10, M.(%)↓=66.9, L.(ms)↓=14622025.11 | 86.7 | — | — | — | — | — | 3.29 | |
| Infinity-8B#Speed=1.0×, #Latency=4.80s, #Param=8.0B2025.12 | 86.6 | 85.1 | — | — | 94.5 | — | — | |
| LongCat-ImageParam.=6B, Resolution=1024 x 10242026.03 | 86.6 | 89.1 | 92.5 | 92 | 93.3 | 87.5 | — | |
| GCPOweighted sampling strategy=false, hybrid inference strategy=Li et al., 2025a2025.10 | 86.6 | — | — | — | — | — | — | |
| Ovis-Image2026.02 | 86.59 | 82.37 | 92.38 | 90.42 | 93.98 | 91.2 | — | |
| InfinityStarType=Hybrid (NSP*), Parameters=8B, Prompt rewriting / self-CoT=true2026.05 | 86.55 | 91.68 | — | — | 91.87 | — | — | |
| UniReasonReasoning capability=w Reasoning2026.02 | 86.21 | 91.78 | 91.23 | 90.76 | 91.12 | 92.27 | — | |
| Show-o2CFG=2.5, Approach=Integrating Autoregressive and Diffusion within Transformers2026.03 | 86.14 | 89 | 91.78 | 89.96 | 91.81 | 91.64 | — | |
| Hunyuan-Image 3.0Reasoning capability=w/o Reasoning2026.02 | 86.1 | 92.12 | 92.53 | 89.13 | 92.13 | 91.92 | — | |
| show-o2Model Category=Unified Multimodal Models2026.01 | 86.1 | — | — | — | — | — | — | |
| HunyuanImage-3.02026.02 | 86.1 | 92.12 | 92.53 | 89.13 | 92.13 | 91.92 | — | |
| HunyuanImage-3.0Type=Unified, Arch.=AR + Diff.2026.04 | 86.1 | 92.12 | 92.53 | 89.13 | 92.13 | 91.92 | — | |
| StageVARBase Model=Infinity-8B, #Speed=2.7×, #Latency=1.77s, #Param=8.0B2025.12 | 86.05 | 85.71 | — | — | 94.43 | — | — | |
| Lumina-DiMOOReasoning capability=w/o Reasoning2026.02 | 86.04 | 81.46 | 92.08 | 88.98 | 94.31 | 82 | — | |
| Lumina-DiMOOType=Unified, Arch.=D-Diff.2026.04 | 86.04 | 81.46 | 92.08 | 88.98 | 94.31 | 82 | — | |
| NextFlowModel Type=AR Model2026.01 | 86 | 92.4 | 90.05 | 90.51 | 92.72 | 91.14 | — | |
| NextFlowType=Unified, Arch.=AR + Diff.2026.04 | 86 | 92.4 | 90.05 | 90.51 | 92.72 | 91.14 | — | |
| ARM-RLType=AR2026.06 | 86 | 90.14 | — | — | 92.08 | — | — | |
| GPT Image 22026.05 | 85.98 | 87.27 | 91.91 | 90.85 | 91.59 | 91.58 | — | |
| HiDream-I1-Full2026.02 | 85.89 | 76.44 | 90.22 | 89.48 | 93.74 | 91.83 | — | |
| HiDream-I1-FullModel Type=Diffusion Model, Variant=Full2026.01 | 85.89 | 76.44 | 90.22 | 89.48 | 93.74 | 91.83 | — | |
| HiDream-I1-FullMode=Full2026.02 | 85.89 | 76.44 | 90.22 | 89.48 | 93.74 | 91.83 | — | |
| HiDream-I1-FullNumber of Parameters=13.5B + 17B2026.05 | 85.89 | 76.44 | 90.22 | 89.48 | 93.74 | 91.83 | — | |
| Recursive Sparse ReasoningM=5, T_latent=5, target layers=multiple layers2026.04 | 85.88 | 88.15 | 90.39 | 88.34 | 91.53 | 82.8 | — | |
| DreamLiteParam.=0.39B, Resolution=1024 x 10242026.03 | 85.8 | 84.9 | 91.6 | 89.7 | 94.3 | 80.4 | — | |
| SD3-mediumSFT=true2026.04 | 85.72 | 85.51 | 89.9 | 88.14 | 92.8 | 80.72 | — | |
| Janus-4oCFG=2.5, Approach=Unifying Understanding and Generation via Next-token Prediction2026.03 | 85.71 | 92.59 | 90.61 | 89.51 | 91.77 | 89.01 | — | |
| SD3-medium2026.04 | 85.65 | 85.71 | 90.24 | 87.39 | 92.15 | 80.4 | — | |
| LineAR (Janus-Pro-7B)Budget Ratio (rho)=1/4, Visual Tokens (N)=576, Throughput=1.58it/s, Speedup=4.49x2025.12 | 85.6 | — | 90.52 | 88.22 | 93.5 | — | — | |
| Janus-Pro-7BBudget Ratio (rho)=Full, Visual Tokens (N)=576, Throughput=0.35it/s, Speedup=1.00x2025.12 | 85.54 | — | 90.34 | 88.28 | 93.23 | — | — | |
| SD3Backbone=SD3-medium2026.02 | 85.43 | 86.32 | 90.71 | 88.24 | 93.35 | 85.6 | — | |
| Recursive Sparse ReasoningM=2, T_latent=2, target layers=multiple layers2026.04 | 85.31 | 87.23 | 89.97 | 87.63 | 91.61 | 84.4 | — | |
| LineAR (Janus-Pro-7B)Budget Ratio (rho)=1/6, Visual Tokens (N)=576, Throughput=1.98it/s, Speedup=5.62x2025.12 | 85.3 | — | 90.32 | 87.97 | 93.38 | — | — | |
| NextStep-1Model Type=AR Model2026.01 | 85.28 | — | — | — | — | — | — | |
| NextStep-1Type=Unidirectional (NPP*), Parameters=14B, Prompt rewriting / self-CoT=true2026.05 | 85.28 | — | — | — | — | — | — | |
| SnapGen++Param.=0.4B, Resolution=1024 x 10242026.03 | 85.2 | — | — | — | — | — | — | |
| InternVL-UType=Unified, Arch.=AR + Diff.2026.04 | 85.18 | 90.39 | 90.78 | 90.68 | 90.29 | 88.77 | — | |
| Dance-GRPOhybrid inference strategy=Li et al., 2025a2025.10 | 85.17 | — | — | — | — | — | — | |
| GPT Image 1Setting=High2026.02 | 85.15 | 88.89 | 88.94 | 89.84 | 92.63 | 90.96 | — | |
| GPT-4o2026.02 | 85.15 | 88.89 | 88.94 | 89.84 | 92.63 | 90.96 | — | |
| GPT Image 1 [High]Model Type=Diffusion Model, Variant=High2026.01 | 85.15 | 88.89 | 88.94 | 89.84 | 92.63 | 90.96 | — | |
| GPT Image 12026.02 | 85.15 | 88.89 | 88.94 | 89.84 | 92.63 | 90.96 | — | |
| GPT Image 1 [High]2026.05 | 85.15 | 88.89 | 88.94 | 89.84 | 92.63 | 90.96 | — | |
| GCPOweighted sampling strategy=true, hybrid inference strategy=Li et al., 2025a2025.10 | 85.14 | — | — | — | — | — | — | |
| BAGELParam.=7B, Resolution=1024 x 10242026.03 | 85.1 | 88.9 | 90.4 | 91.3 | 90.8 | 88.7 | — | |
| BAGELReasoning capability=w Reasoning2026.02 | 85.07 | 88.94 | 90.37 | 91.29 | 90.82 | 88.67 | — | |
| BAGELCFG=2.5, Approach=Integrating Autoregressive and Diffusion within Transformers2026.03 | 85.07 | 88.94 | 90.37 | 91.29 | 90.82 | 88.67 | — | |
| BAGELType=Unified, Arch.=AR + Diff.2026.04 | 85.07 | 88.94 | 90.37 | 91.29 | 90.82 | 88.67 | — | |
| Flow-GRPOhybrid inference strategy=Li et al., 2025a2025.10 | 85.05 | — | — | — | — | — | — | |
| Janus-Pro-1BBudget Ratio (rho)=Full, Visual Tokens (N)=576, Throughput=0.89it/s, Speedup=1.00x2025.12 | 85 | — | 90.4 | 88.78 | 92.8 | — | — | |
| Z-Image-TurboMode=Turbo2026.02 | 84.86 | 91.29 | 89.59 | 90.14 | 92.16 | 88.68 | — | |
| Z-Image-TurboType=Gen. Only, Arch.=Diff.2026.04 | 84.86 | 91.29 | 89.59 | 90.14 | 92.16 | 88.68 | — | |
| Z-Image-TurboNumber of Parameters=4B + 6B2026.05 | 84.86 | 91.29 | 89.59 | 90.14 | 92.16 | 88.68 | — | |
| SANA-1.6BParam.=1.6B, Resolution=1024 x 10242026.03 | 84.8 | 86 | 91.5 | 88.9 | 91.9 | 90.7 | — | |
| FLUX-schnellType=Bidirectional (Diff.), Parameters=12B, Prompt rewriting / self-CoT=false2026.05 | 84.8 | — | — | — | — | — | — | |
| SANAType=Bidirectional (Diff.), Parameters=1.6B, Prompt rewriting / self-CoT=false2026.05 | 84.8 | — | — | — | 91.9 | — | — | |
| LineAR (Janus-Pro-1B)Budget Ratio (rho)=1/4, Visual Tokens (N)=576, Throughput=4.03it/s, Speedup=4.53x2025.12 | 84.79 | — | 90.26 | 88.26 | 93.19 | — | — | |
| Prompt ReinjectionBackbone=FLUX.1-Dev2026.02 | 84.67 | 84.8 | 90.26 | 88.16 | 93.62 | 82.8 | — | |
| DeepGen1.0Param.=2B, Resolution=1024 x 10242026.03 | 84.6 | 84.2 | 91 | 88.6 | 94 | 85.3 | — |