Text-to-image generation on MS-COCO 2014 (val)
2.47FIDRandom Train Images
Evaluation Results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Random Train ImagesEvaluation Protocol=Zero-shot, Guidance Scale=1.22022.06 | 2.47 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PartiModel Type=Autoregressive, Evaluation Protocol=Finetuned, Guidance Scale=1.22022.06 | 3.22 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RAPHAELData distribution=English, Resolution=256 × 256, Evaluation protocol=zero-shot, Sample size=30k2023.09 | 6.61 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ERNIE-ViLG 2.0Architecture=Non-AR based, Parameters=24B2023.03 | 6.75 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ERNIE-ViLG 2.0Data distribution=English and Chinese, Resolution=256 × 256, Evaluation protocol=zero-shot, Sample size=30k2023.09 | 6.75 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Retrieval BaselineEvaluation Protocol=Finetuned, Guidance Scale=1.22022.06 | 6.82 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Re-Imagenzero-shot=true, num_samples=300002023.05 | 6.88 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| eDiff-IData distribution=English, Resolution=256 × 256, Evaluation protocol=zero-shot, Sample size=30k2023.09 | 6.95 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PartiModel Type=Autoregressive, Evaluation Protocol=Zero-shot, Guidance Scale=1.22022.06 | 7.23 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Partizero-shot=true, num_samples=300002023.05 | 7.23 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ImagenModel Type=Diffusion, Evaluation Protocol=Zero-shot, Guidance Scale=1.22022.06 | 7.27 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ImagenArchitecture=Non-AR based, Parameters=4.9B2023.03 | 7.27 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Imagenzero-shot=true, num_samples=300002023.05 | 7.27 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Make-A-SceneModel Type=Autoregressive, Evaluation Protocol=Finetuned, Guidance Scale=1.22022.06 | 7.55 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LAFITE2zero-shot=true, num_samples=300002023.05 | 8.42 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Stable DiffusionData distribution=English, Resolution=256 × 256, Evaluation protocol=zero-shot, Sample size=30k2023.09 | 8.59 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SD v1.5zero-shot=true, num_samples=30000, classifier-free guidance scaling factor=3, DDIM inference steps=2502023.05 | 9.22 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| XMC-GANModel Type=GAN, Evaluation Protocol=Finetuned, Guidance Scale=1.22022.06 | 9.33 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BDMData distribution=Chinese, Resolution=256 × 256, Evaluation protocol=zero-shot, Sample size=30k, classifier-free guidance scale=3.7, Chinese language branch weight=1.5, backbone=SD1.52023.09 | 9.93 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DALL-E 2Model Type=Diffusion, Evaluation Protocol=Zero-shot, Guidance Scale=1.22022.06 | 10.39 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DALL-E-2Architecture=Non-AR based, Parameters=6.5B2023.03 | 10.39 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DALL-E 2zero-shot=true, num_samples=300002023.05 | 10.39 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DALL-E 2Data distribution=English, Resolution=256 × 256, Evaluation protocol=zero-shot, Sample size=30k2023.09 | 10.39 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAGVLT (T2I only)Architecture=Non-AR based, Parameters=447M, Modality=T2I only2023.03 | 10.74 | — | — | 23.94 | 8.12 | — | — | — | — | — | — | — | — | — | |
| DALL-E 2zero-shot=true2023.04 | 10.9 | — | — | — | — | 1 | 0.5387 | 1 | — | — | — | — | — | — | |
| Make-A-SceneModel Type=Autoregressive, Evaluation Protocol=Zero-shot, Guidance Scale=1.22022.06 | 11.84 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Make-A-SceneArchitecture=AR based, Parameters=4B2023.03 | 11.84 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Make-A-Scenezero-shot=true, num_samples=300002023.05 | 11.84 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAGVLTParameters=447M, Modality=Available for both T2I & I2T2023.03 | 12.08 | — | — | 22.75 | 8.12 | — | — | — | — | — | — | — | — | — | |
| GILLzero-shot=true, num_samples=30000, classifier-free guidance scaling factor=3, DDIM inference steps=2502023.05 | 12.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GLIDEModel Type=Diffusion, Evaluation Protocol=Zero-shot, Guidance Scale=1.22022.06 | 12.24 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GLIDEArchitecture=Non-AR based, Parameters=3.5B2023.03 | 12.24 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GLIDEzero-shot=true, num_samples=300002023.05 | 12.24 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GLIDEData distribution=English, Resolution=256 × 256, Evaluation protocol=zero-shot, Sample size=30k2023.09 | 12.24 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LDMData distribution=English, Resolution=256 × 256, Evaluation protocol=zero-shot, Sample size=30k2023.09 | 12.61 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Parti-350MArchitecture=AR based, Parameters=350M2023.03 | 14.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ERNIE-ViLGData distribution=English and Chinese, Resolution=256 × 256, Evaluation protocol=zero-shot, Sample size=30k2023.09 | 14.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DPM-Solver-v3NFE=12, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 15.13 | — | 0.265 | — | — | — | — | — | — | — | — | — | — | — | |
| DPM-Solver-v3NFE=20, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 15.23 | — | 0.265 | — | — | — | — | — | — | — | — | — | — | — | |
| DPM-Solver-v3NFE=15, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 15.3 | — | 0.265 | — | — | — | — | — | — | — | — | — | — | — | |
| DPM-Solver-v3NFE=10, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 15.32 | — | 0.265 | — | — | — | — | — | — | — | — | — | — | — | |
| DPM-Solver-v3NFE=8, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 15.41 | — | 0.264 | — | — | — | — | — | — | — | — | — | — | — | |
| DPM-Solver++NFE=20, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 15.72 | — | 0.265 | — | — | — | — | — | — | — | — | — | — | — | |
| DPM-Solver++NFE=12, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 15.78 | — | 0.266 | — | — | — | — | — | — | — | — | — | — | — | |
| DPM-Solver++NFE=15, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 15.84 | — | 0.265 | — | — | — | — | — | — | — | — | — | — | — | |
| DPM-Solver++NFE=10, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 15.93 | — | 0.265 | — | — | — | — | — | — | — | — | — | — | — | |
| UniPCNFE=20, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 15.94 | — | 0.265 | — | — | — | — | — | — | — | — | — | — | — | |
| UniPCNFE=15, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 16.06 | — | 0.265 | — | — | — | — | — | — | — | — | — | — | — | |
| UniPCNFE=12, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 16.09 | — | 0.265 | — | — | — | — | — | — | — | — | — | — | — | |
| UniPCNFE=10, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 16.15 | — | 0.265 | — | — | — | — | — | — | — | — | — | — | — | |
| UniPCNFE=8, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 16.2 | — | 0.265 | — | — | — | — | — | — | — | — | — | — | — | |
| DPM-Solver++NFE=8, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 16.4 | — | 0.265 | — | — | — | — | — | — | — | — | — | — | — | |
| DPM-Solver-v3NFE=6, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 16.41 | — | 0.262 | — | — | — | — | — | — | — | — | — | — | — | |
| ARGVLTParameters=447M, Modality=Available for both T2I & I2T2023.03 | 16.93 | — | — | 22.5 | 1 | — | — | — | — | — | — | — | — | — | |
| UniPCNFE=6, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 17.32 | — | 0.263 | — | — | — | — | — | — | — | — | — | — | — | |
| DPM-Solver++NFE=6, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 17.44 | — | 0.265 | — | — | — | — | — | — | — | — | — | — | — | |
| CogView2Model Type=Autoregressive, Evaluation Protocol=Finetuned, Guidance Scale=1.22022.06 | 17.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Stable Diffusion 1.4Resolution=256x256, zero-shot=true2023.04 | 17.9 | — | — | — | — | 4 | -0.0857 | 2 | — | — | — | — | — | — | |
| Retrieval BaselineEvaluation Protocol=Zero-shot, Guidance Scale=1.22022.06 | 17.97 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Versatile DiffusionResolution=256x256, zero-shot=true2023.04 | 18.4 | — | — | — | — | 5 | -0.5485 | 3 | — | — | — | — | — | — | |
| UniPCNFE=5, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 18.77 | — | 0.262 | — | — | — | — | — | — | — | — | — | — | — | |
| Stable Diffusion 2.1-baseResolution=256x256, zero-shot=true2023.04 | 18.8 | — | — | — | — | 2 | 0.1553 | 4 | — | — | — | — | — | — | |
| DPM-Solver-v3NFE=5, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 18.83 | — | 0.26 | — | — | — | — | — | — | — | — | — | — | — | |
| DPM-Solver++NFE=5, Guidance scale=7.5, Model=Stable-Diffusion [43]2023.10 | 18.87 | — | 0.263 | — | — | — | — | — | — | — | — | — | — | — | |
| DeepCacheN=22025.12 | 19.44 | — | — | — | — | — | 1.07 | — | 22.85 | 0.24 | 1.29 | — | — | — | |
| ∆-DiTb=302025.12 | 19.72 | — | — | — | — | — | 1.06 | — | 22.85 | 0.25 | 1.62 | — | — | — | |
| Stable Diffusion 3.5 LargeSampling steps=502025.12 | 20.01 | — | — | — | — | — | 1.09 | — | 22.9 | — | — | — | — | — | |
| BlockDanceN=32025.12 | 20.26 | — | — | — | — | — | 1.08 | — | 22.89 | 0.04 | 1.49 | — | — | — | |
| OpenjourneyResolution=256x256, zero-shot=true2023.04 | 20.7 | — | — | — | — | 3 | -0.0455 | 5 | — | — | — | — | — | — | |
| CorGi+D=5, γ=30, δ=22025.12 | 20.7 | — | — | — | — | — | 1.05 | — | 22.76 | 0.08 | 2.02 | — | — | — | |
| CorGi+D=6, γ=28, δ=22025.12 | 20.71 | — | — | — | — | — | 1.04 | — | 22.74 | 0.09 | 2.02 | — | — | — | |
| CorGiD=5, γ=30, δ=22025.12 | 21.38 | — | — | — | — | — | 1.02 | — | 22.74 | 0.08 | 2.09 | — | — | — | |
| CorGiD=6, γ=28, δ=22025.12 | 21.45 | — | — | — | — | — | 1.02 | — | 22.73 | 0.09 | 2.08 | — | — | — | |
| RASratio=30%2025.12 | 21.5 | — | — | — | — | — | 1 | — | 22.45 | 0.11 | 1.99 | — | — | — | |
| ARGVLT (T2I only)Architecture=AR based, Parameters=447M, Modality=T2I only2023.03 | 21.8 | — | — | 19.27 | 1 | — | — | — | — | — | — | — | — | — | |
| CogView2Model Type=Autoregressive, Evaluation Protocol=Zero-shot, Guidance Scale=1.22022.06 | 24 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CogView2Architecture=AR based, Parameters=6B2023.03 | 24 | — | — | 22.4 | — | — | — | — | — | — | — | — | — | — | |
| SD-1.5Resolution=512x5122025.09 | 24.01 | — | 30.17 | — | — | — | — | — | — | — | — | — | — | — | |
| SD-2.1-baseResolution=512x5122025.09 | 24.25 | — | 31.32 | — | — | — | — | — | — | — | — | — | — | — | |
| MaskAttn-SDXLSetting=mid-res, default, Resolution=1024x10242025.09 | 24.57 | — | 31.75 | — | — | — | — | — | — | — | — | — | — | — | |
| FireFlowSteps=10, NFE=11, ODE Solver=2nd-order2024.12 | 25.16 | — | 31.42 | — | — | — | — | — | — | — | — | — | — | — | |
| PixArt-ΣInference Setting=method-recommended2025.09 | 25.23 | — | 31.55 | — | — | — | — | — | — | — | — | — | — | — | |
| PixArt-αInference Setting=method-recommended2025.09 | 25.68 | — | 31.49 | — | — | — | — | — | — | — | — | — | — | — | |
| SDXLResolution=1024x10242025.09 | 25.77 | — | 31.53 | — | — | — | — | — | — | — | — | — | — | — | |
| RF-SolverSteps=10, NFE=20, ODE Solver=2nd-order2024.12 | 25.93 | — | 31.35 | — | — | — | — | — | — | — | — | — | — | — | |
| CogView 2Resolution=256x256, zero-shot=true2023.04 | 26.2 | — | — | — | — | 6 | -0.851 | 6 | — | — | — | — | — | — | |
| DeepCacheBlock-segment caching (N)=2, Speedup=1.26×2025.12 | 26.24 | — | — | — | — | 1.09 | — | — | 23.11 | 0.25 | — | — | — | — | |
| BlockDanceBlock-segment caching (N)=3, Speedup=1.50×2025.12 | 26.24 | — | — | — | — | 1.1 | — | — | 23.12 | 0.05 | — | — | — | — | |
| FLUX-devSteps=20, NFE=20, ODE Solver=1st-order2024.12 | 26.77 | — | 31.44 | — | — | — | — | — | — | — | — | — | — | — | |
| CogViewModel Type=Autoregressive, Evaluation Protocol=Zero-shot, Guidance Scale=1.22022.06 | 27.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CogViewArchitecture=AR based, Parameters=4B2023.03 | 27.1 | — | — | 18.2 | — | — | — | — | — | — | — | — | — | — | |
| OBS-DiffBase Model=SD 3-medium, Sparsity (%)=502025.10 | 27.2 | — | 0.3167 | — | — | — | 0.6468 | — | — | — | — | — | — | — | |
| OBS-DiffBase Model=SD v2.1-base, Sparsity (%)=502025.10 | 27.41 | — | 0.3102 | — | — | — | -0.0356 | — | — | — | — | — | — | — | |
| CorGi+Caching parameters (D, γ, δ)=D=6, γ=47, δ=2, Speedup=1.98×2025.12 | 27.48 | — | — | — | — | 1.01 | — | — | 22.82 | 0.12 | — | — | — | — | |
| DALL-EArchitecture=AR based, Parameters=12B2023.03 | 27.5 | — | — | 17.9 | — | — | — | — | — | — | — | — | — | — | |
| CorGi+Caching parameters (D, γ, δ)=D=5, γ=49, δ=2, Speedup=1.91×2025.12 | 27.56 | — | — | — | — | 1.05 | — | — | 22.95 | 0.09 | — | — | — | — | |
| FLUX.1-devInference steps=302025.12 | 27.77 | — | — | — | — | 1.08 | — | — | 23.11 | — | — | — | — | — | |
| MagnitudeBase Model=SD v2.1-base, Sparsity (%)=402025.10 | 27.86 | — | 0.3111 | — | — | — | 0.1864 | — | — | — | — | — | — | — | |
| WandaBase Model=SD v2.1-base, Sparsity (%)=402025.10 | 27.96 | — | 0.3122 | — | — | — | 0.1367 | — | — | — | — | — | — | — | |
| DALL-EModel Type=Autoregressive, Evaluation Protocol=Zero-shot, Guidance Scale=1.22022.06 | 28 | — | — | — | — | — | — | — | — | — | — | — | — | — |