Text-to-Image Generation on MS COCO zero-shot
6.95FIDeDiff-I
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| eDiff-IRes.=256, Time (s)=32.0s, # Steps=25+10, # Param.=9B, category=Diffusion Models2024.06 | 6.95 | — | — | — | — | — | — | |
| Parti-20BRes.=256, # Param.=20B, category=Autoregressive Models2024.06 | 7.23 | — | — | — | — | — | — | |
| ImagenRes.=256, Time (s)=9.1s, # Param.=3B, category=Diffusion Models2024.06 | 7.27 | — | — | — | — | — | — | |
| MuseRes.=256, Time (s)=1.3s, # Steps=24, # Param.=3B, category=Masked Models2024.06 | 7.88 | 0.32 | — | — | — | — | — | |
| Parti-3BRes.=256, Time (s)=6.4s, # Param.=3B, category=Autoregressive Models2024.06 | 8.1 | — | — | — | — | — | — | |
| SiD-LSGRes.=512, Time (s)=0.09s, # Steps=1, # Param.=0.9B, kappa=1.5, training_time=double, base_model=SD1.52024.06 | 8.15 | 0.304 | — | — | — | — | — | |
| SiD-LSGRes.=512, Time (s)=0.09s, # Steps=1, # Param.=0.9B, kappa=1.5, base_model=SD1.52024.06 | 8.71 | 0.302 | — | — | — | — | — | |
| PixArt-αcfg=4.5, Backbone=PixArt-α-based backbone, #Params=610.86M, NFEs=20, Type=DiT, Trainable params=610.86M, Image-free?=✗, Training Data=25M2026.03 | 8.75 | 0.32 | — | 0.75 | 0.45 | — | — | |
| SD1.5 (CFG=3)Res.=512, Time (s)=2.59s, # Steps=50, # Param.=0.9B, CFG=32024.06 | 8.78 | — | — | — | — | — | — | |
| SD 1.5cfg=3.0, Backbone=Stable Diffusion 1.5, #Params=860M, NFEs=25, Type=U-Net, Trainable params=860M, Image-free?=✗, Training Data=5B2026.03 | 8.78 | 0.3 | — | 0.59 | 0.53 | — | — | |
| GigaGANRes.=512, Time (s)=0.13s, # Steps=1, # Param.=1B, category=Generative Adversarial Networks (GANs)2024.06 | 9.09 | — | — | — | — | — | — | |
| SiD-LSGRes.=512, Time (s)=0.09s, # Steps=1, # Param.=0.9B, kappa=1.5, base_model=SD2.12024.06 | 9.52 | 0.308 | — | — | — | — | — | |
| SiD-LSGRes.=512, Time (s)=0.09s, # Steps=1, # Param.=0.9B, kappa=2, base_model=SD1.52024.06 | 9.56 | 0.313 | — | — | — | — | — | |
| SD 2.1cfg=3.0, Backbone=Stable Diffusion 2.1, #Params=865M, NFEs=1, Type=U-Net, Trainable params=865M, Image-free?=✗, Training Data=5B2026.03 | 9.6 | 0.32 | — | 0.59 | 0.5 | — | — | |
| DALL-E 2Res.=256, # Steps=250+27, # Param.=5.5B, category=Diffusion Models2024.06 | 10.39 | — | — | — | — | — | — | |
| Parti-750MRes.=256, # Param.=750M, category=Autoregressive Models2024.06 | 10.71 | — | — | — | — | — | — | |
| WaDiBackbone=Stable Diffusion 1.5, #Params=860M, NFEs=1, Type=LoRaD, Trainable params=83.80M, Image-free?=✓, Training Data=1.4M2026.03 | 10.79 | 0.31 | — | 0.62 | 0.48 | — | — | |
| SiD-LSGRes.=512, Time (s)=0.09s, # Steps=1, # Param.=0.9B, kappa=2, base_model=SD2.12024.06 | 10.97 | 0.318 | — | — | — | — | — | |
| DMD (CFG=3)Res.=512, Time (s)=0.09s, # Steps=1, # Param.=0.9B, CFG=32024.06 | 11.49 | — | — | — | — | — | — | |
| DMDBackbone=Stable Diffusion 1.5, #Params=860M, NFEs=1, Type=U-Net, Trainable params=860M, Image-free?=✗, Training Data=3M2026.03 | 11.49 | 0.32 | — | — | — | — | — | |
| InstaFlow-1.7BRes.=512, Time (s)=0.12s, # Steps=1, # Param.=1.7B2024.06 | 11.83 | — | — | — | — | — | — | |
| Make-A-SceneRes.=256, Time (s)=25.0s, category=Autoregressive Models2024.06 | 11.84 | — | — | — | — | — | — | |
| GLIDERes.=256, Time (s)=15.0s, # Steps=250+27, # Param.=5B, category=Diffusion Models2024.06 | 12.24 | — | — | — | — | — | — | |
| WaDiBackbone=Stable Diffusion 2.1, #Params=865M, NFEs=1, Type=LoRaD, Trainable params=94.43M, Image-free?=✓, Training Data=1.4M2026.03 | 12.34 | 0.31 | — | 0.6 | 0.48 | — | — | |
| LDM-KL-8-G*Res.=256, Time (s)=3.7s, # Steps=250, # Param.=1.45B, category=Diffusion Models2024.06 | 12.63 | — | — | — | — | — | — | |
| UFOGenRes.=512, Time (s)=0.09s, # Steps=1, # Param.=0.9B2024.06 | 12.78 | — | — | — | — | — | — | |
| UFOGenBackbone=Stable Diffusion 1.5, #Params=860M, NFEs=1, Type=U-Net, Trainable params=860M, Image-free?=✗, Training Data=12M2026.03 | 12.78 | — | — | — | — | — | — | |
| DMD2Backbone=Stable Diffusion 1.5, #Params=860M, NFEs=1, Type=U-Net, Trainable params=860M, Image-free?=✓, Training Data=1.4M, reproduced=true2026.03 | 12.96 | 0.3 | — | 0.6 | 0.47 | — | — | |
| InstaFlow-0.9BRes.=512, Time (s)=0.09s, # Steps=1, # Param.=0.9B2024.06 | 13.1 | 0.28 | — | — | — | — | — | |
| InstaFlowBackbone=Stable Diffusion 1.5, #Params=860M, NFEs=1, Type=U-Net, Trainable params=860M, Image-free?=✗, Training Data=3.2M2026.03 | 13.1 | 0.28 | — | 0.53 | 0.45 | — | — | |
| SiD-LSGRes.=512, Time (s)=0.09s, # Steps=1, # Param.=0.9B, kappa=3, base_model=SD1.52024.06 | 13.21 | 0.314 | — | — | — | — | — | |
| SD1.5 (CFG=8)Res.=512, Time (s)=2.59s, # Steps=50, # Param.=0.9B, CFG=82024.06 | 13.45 | 0.322 | — | — | — | — | — | |
| SD2.1 baseRes.=512, Time (s)=0.77s, # Steps=25, # Param.=0.9B2024.06 | 13.45 | 0.3 | — | — | — | — | — | |
| TiUEBackbone=Stable Diffusion 2.1, #Params=865M, NFEs=1, Type=U-Net, Trainable params=865M, Image-free?=✓, Training Data=1.4M2026.03 | 13.49 | 0.31 | — | 0.59 | 0.48 | — | — | |
| SiD-LSGRes.=512, Time (s)=0.09s, # Steps=1, # Param.=0.9B, kappa=3, base_model=SD2.12024.06 | 13.5 | 0.321 | — | — | — | — | — | |
| StyleGAN-TRes.=512, Time (s)=0.10s, # Steps=1, # Param.=1B, category=Generative Adversarial Networks (GANs)2024.06 | 13.9 | 0.293 | — | — | — | — | — | |
| SiD-LSGBackbone=Stable Diffusion 1.5, #Params=860M, NFEs=1, Type=U-Net, Trainable params=860M, Image-free?=✓, Training Data=1.4M, reproduced=true2026.03 | 14.27 | 0.3 | — | 0.56 | 0.48 | — | — | |
| DMD (CFG=8)Res.=512, Time (s)=0.09s, # Steps=1, # Param.=0.9B, CFG=82024.06 | 14.93 | 0.32 | — | — | — | — | — | |
| SiD-LSGBackbone=Stable Diffusion 2.1, #Params=865M, NFEs=1, Type=U-Net, Trainable params=865M, Image-free?=✓, Training Data=1.4M, reproduced=true2026.03 | 15.17 | 0.3 | — | 0.56 | 0.46 | — | — | |
| Swiftbrushv2Backbone=Stable Diffusion 2.1, #Params=865M, NFEs=1, Type=U-Net+LoRA, Trainable params=884.14M, Image-free?=✓, Training Data=1.4M, reproduced=true2026.03 | 15.98 | 0.33 | — | 0.58 | 0.47 | — | — | |
| SD-TurboBackbone=Stable Diffusion 2.1, #Params=865M, NFEs=1, Type=U-Net, Trainable params=865M, Image-free?=✗2026.03 | 16.14 | 0.33 | — | 0.65 | 0.35 | — | — | |
| ADD (SD-Turbo)Res.=512, # Steps=1, category=Distilled Stable Diffusion 2.12024.06 | 16.25 | 0.335 | — | — | — | — | — | |
| SiD-LSGRes.=512, Time (s)=0.09s, # Steps=1, # Param.=0.9B, kappa=4.5, base_model=SD2.12024.06 | 16.54 | 0.322 | — | — | — | — | — | |
| SiD-LSGRes.=512, Time (s)=0.09s, # Steps=1, # Param.=0.9B, kappa=4.5, base_model=SD1.52024.06 | 16.59 | 0.317 | — | — | — | — | — | |
| SwiftBrushRes.=512, Time (s)=0.09s, # Steps=1, # Param.=0.9B2024.06 | 16.67 | 0.29 | — | — | — | — | — | |
| SwiftbrushBackbone=Stable Diffusion 2.1, #Params=865M, NFEs=1, Type=U-Net, Trainable params=865M, Image-free?=✓, Training Data=1.4M2026.03 | 16.67 | 0.29 | — | 0.47 | 0.46 | — | — | |
| PCMBackbone=Stable Diffusion 1.5, #Params=860M, NFEs=1, Type=U-Net, Trainable params=860M, Image-free?=✗, Training Data=3M2026.03 | 17.91 | 0.29 | — | — | — | — | — | |
| WaDiBackbone=PixArt-α-based backbone, #Params=610.86M, NFEs=1, Type=LoRaD, Trainable params=81.22M, Image-free?=✓, Training Data=1.4M2026.03 | 18.99 | 0.3 | — | 0.64 | 0.29 | — | — | |
| ADD (SDXL-Turbo)Res.=512, # Steps=1, category=Distilled Stable Diffusion XL2024.06 | 19.08 | 0.343 | — | — | — | — | — | |
| UniPCRes.=512, Time (s)=0.26s, # Steps=4, # Param.=0.9B2024.06 | 22.3 | 0.31 | — | — | — | — | — | |
| DPM++Res.=512, Time (s)=0.26s, # Steps=4, # Param.=0.9B2024.06 | 22.44 | 0.31 | — | — | — | — | — | |
| Hyper-SDBackbone=Stable Diffusion 1.5, #Params=860M, NFEs=1, Type=LoRA, Trainable params=67.25M, Image-free?=✗2026.03 | 22.9 | 0.31 | — | 0.62 | 0.25 | — | — | |
| LCM-LoRARes.=512, Time (s)=0.19s, # Steps=4, # Param.=0.9B2024.06 | 23.62 | 0.3 | — | — | — | — | — | |
| YOSOBackbone=Stable Diffusion 1.5, #Params=860M, NFEs=1, Type=LoRA, Trainable params=67.25M, Image-free?=✗, Training Data=4M2026.03 | 23.68 | 0.29 | — | 0.56 | 0.36 | — | — | |
| CogView2Res.=256, # Param.=6B, category=Autoregressive Models2024.06 | 24 | — | — | — | — | — | — | |
| PG-SBBackbone=PixArt-α-based backbone, #Params=610.86M, NFEs=1, Type=DiT, Trainable params=610.86M, Image-free?=✓, Training Data=1.4M, reproduced=true2026.03 | 25.58 | 0.28 | — | 0.53 | 0.27 | — | — | |
| LAFITERes.=256, Time (s)=0.02s, # Steps=1, # Param.=75M, category=Generative Adversarial Networks (GANs)2024.06 | 26.94 | — | — | — | — | — | — | |
| DALL-ERes.=256, # Param.=12B, category=Autoregressive Models2024.06 | 27.5 | — | — | — | — | — | — | |
| SwiftbrushBackbone=PixArt-α-based backbone, #Params=610.86M, NFEs=1, Type=DiT, Trainable params=610.86M, Image-free?=✓, Training Data=1.4M, reproduced=true2026.03 | 29.89 | 0.28 | — | 0.5 | 0.26 | — | — | |
| Guided DistillationRes.=512, # Steps=1, # Param.=0.9B2024.06 | 37.3 | 0.27 | — | — | — | — | — | |
| BOOTRes.=512, Time (s)=0.09s, # Steps=1, # Param.=0.9B2024.06 | 48.2 | 0.26 | — | — | — | — | — | |
| LCM-LoRABackbone=Stable Diffusion 1.5, #Params=860M, NFEs=1, Type=LoRA, Trainable params=67.50M, Image-free?=✗, Training Data=12M2026.03 | 77.73 | 0.24 | — | 0.22 | 0.15 | — | — | |
| LCM-LoRARes.=512, Time (s)=0.07s, # Steps=1, # Param.=0.9B2024.06 | 77.9 | 0.24 | — | — | — | — | — | |
| SD2.1-baseRes.=512, Time (s)=0.09s, # Steps=1, # Param.=0.9B2024.06 | 202.14 | 0.08 | — | — | — | — | — | |
| 2-RFNFE=12024.12 | — | 0.271 | — | — | — | 47 | 0.09 | |
| 2-RFNFE=22024.12 | — | 0.296 | — | — | — | 31.3 | 0.13 | |
| 2-RFNFE=12024.12 | — | 0.282 | 36.83 | 0.3 | 0.33 | — | — | |
| 2-RF (teacher)NFE=252024.12 | — | 0.315 | — | — | — | 21.5 | 0.88 | |
| 2-RF (teacher)NFE=252024.12 | — | — | 11.08 | — | — | — | — | |
| CrossFlow#Params.=0.95B, zero-shot=true2024.12 | — | — | 9.63 | — | — | — | — | |
| CrossFlow (Sin-Cos)#Params.=0.95B, zero-shot=true2024.12 | — | — | 8.95 | — | — | — | — | |
| DALL-E#Params.=12.0B, zero-shot=true2024.12 | — | — | 27.5 | — | — | — | — | |
| DALL-E 2#Params.=6.5B, zero-shot=true2024.12 | — | — | 10.39 | — | — | — | — | |
| GLIDE#Params.=5.0B, zero-shot=true2024.12 | — | — | 12.24 | — | — | — | — | |
| Guided DistillNFE=12024.12 | — | 0.27 | 37.3 | — | — | — | — | |
| Guided Distill.NFE=12024.12 | — | 0.275 | — | — | — | 37.2 | 0.09 | |
| Guided Distill.NFE=22024.12 | — | 0.297 | — | — | — | 26 | 0.13 | |
| Imagen#Params.=3.0B, zero-shot=true2024.12 | — | — | 7.27 | — | — | — | — | |
| InstaFlow-0.9BNFE=12024.12 | — | 0.304 | — | — | — | 23.4 | 0.09 | |
| InstaFlow-0.9BNFE=12024.12 | — | 0.28 | 13.1 | 0.53 | 0.45 | — | — | |
| InstaFlow-1.7BNFE=12024.12 | — | 0.309 | — | — | — | 22.4 | 0.12 | |
| InstaFlow-1.7BNFE=12024.12 | — | — | 11.83 | — | — | — | — | |
| LCMNFE=12024.12 | — | 0.24 | 35.56 | — | — | — | — | |
| LDM#Params.=1.4B, zero-shot=true2024.12 | — | — | 12.63 | — | — | — | — | |
| LDMv1.5#Params.=0.9B, zero-shot=true2024.12 | — | — | 9.62 | — | — | — | — | |
| Ours-0.9BNFE=12024.12 | — | 0.313 | — | — | — | 22.09 | 0.09 | |
| Ours-0.9BNFE=22024.12 | — | 0.317 | — | — | — | 21.2 | 0.13 | |
| Ours-0.9BNFE=12024.12 | — | 0.312 | 11.91 | 0.54 | 0.47 | — | — | |
| Ours-0.9BNFE=22024.12 | — | 0.315 | 11.46 | 0.57 | 0.46 | — | — | |
| PixArt-α#Params.=0.6B, zero-shot=true2024.12 | — | — | 7.32 | — | — | — | — | |
| RAPHAEL#Params.=3.0B, zero-shot=true2024.12 | — | — | 6.61 | — | — | — | — | |
| SD TurboNFE=12024.12 | — | 0.33 | 16.1 | 0.65 | 0.35 | — | — | |
| UFOGenNFE=12024.12 | — | 0.311 | — | — | — | 22.5 | 0.09 | |
| UFOGenNFE=12024.12 | — | — | 12.78 | — | — | — | — |