Text-to-Image Generation on COCO 2014 (val)
7.11FIDCTEC + Composer
Evaluation Results
| Method | Links | ||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| CTEC + ComposerBitwidth (W/A)=2/82026.03 | 7.11 | — | — | — | — | — | 82.5 | — | — | — | — | — | — | — | — | — | 191.74 | 7.45 | — | — | — | — | — | — | — | — | |
| CTECBitwidth (W/A)=2/82026.03 | 7.43 | — | — | — | — | — | 80.2 | — | — | — | — | — | — | — | — | — | 176.37 | 7.98 | — | — | — | — | — | — | — | — | |
| CTEC + ComposerBitwidth (W/A)=4/82026.03 | 8.25 | — | — | — | — | — | 94.15 | — | — | — | — | — | — | — | — | — | 359.2 | 7.11 | — | — | — | — | — | — | — | — | |
| CTECBitwidth (W/A)=4/82026.03 | 8.52 | — | — | — | — | — | 93.81 | — | — | — | — | — | — | — | — | — | 355.62 | 7.31 | — | — | — | — | — | — | — | — | |
| Q-Diffusion + ComposerBitwidth (W/A)=4/82026.03 | 8.95 | — | — | — | — | — | 92.35 | — | — | — | — | — | — | — | — | — | 347.4 | 8.92 | — | — | — | — | — | — | — | — | |
| Q-DiffusionBitwidth (W/A)=4/82026.03 | 9.29 | — | — | — | — | — | 91.06 | — | — | — | — | — | — | — | — | — | 336.8 | 9.29 | — | — | — | — | — | — | — | — | |
| DPMSolverSteps=252026.06 | 9.78 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DDIM (teacher)Steps=322026.06 | 10.05 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MA-ReflowSteps=4, K=42026.06 | 10.66 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Full PrecisionBitwidth (W/A)=32/322026.03 | 11.28 | — | — | — | — | — | 93.66 | — | — | — | — | — | — | — | — | — | 364.73 | 7.7 | — | — | — | — | — | — | — | — | |
| PCMSteps=42026.06 | 11.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PeRFlowSteps=42026.06 | 12.01 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Flash DiffusionSteps=42026.06 | 12.41 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| 2-ReFlowSteps=42026.06 | 15.32 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DMD2EMPURPLE=Yes, Cosine similarity=No, Resolution=1024 x 10242026.07 | 15.84 | — | — | — | — | — | — | — | — | — | — | — | 30.81 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DMD2EMPURPLE=Yes, Cosine similarity=Yes, Resolution=1024 x 10242026.07 | 16.05 | — | — | — | — | — | — | — | — | — | — | — | 30.87 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SDXL-LightningEMPURPLE=Yes, Cosine similarity=Yes, Resolution=1024 x 10242026.07 | 18.21 | — | — | — | — | — | — | — | — | — | — | — | 29.55 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CADSBackbone=SD3 (MMDiT), Guidance=CADS2025.12 | 18.32 | — | — | — | — | — | 69.9 | 31.6 | — | — | — | 28.36 | 26.39 | 1.219 | 0.644 | — | — | — | — | — | — | — | — | — | — | — | |
| SDXL-LightningEMPURPLE=Yes, Cosine similarity=No, Resolution=1024 x 10242026.07 | 18.55 | — | — | — | — | — | — | — | — | — | — | — | 29.49 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SAGBackbone=SD3 (MMDiT), Guidance=SAG2025.12 | 18.74 | — | — | — | — | — | 49.6 | 40 | — | — | — | 28.4 | 24.16 | 0.588 | 0.517 | — | — | — | — | — | — | — | — | — | — | — | |
| Flash SDXLEMPURPLE=Yes, Cosine similarity=No, Resolution=1024 x 10242026.07 | 18.95 | — | — | — | — | — | — | — | — | — | — | — | 29.52 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Flash SDXLEMPURPLE=Yes, Cosine similarity=Yes, Resolution=1024 x 10242026.07 | 19.1 | — | — | — | — | — | — | — | — | — | — | — | 29.49 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DMD2EMPURPLE=No, Cosine similarity=–, Resolution=1024 x 10242026.07 | 19.35 | — | — | — | — | — | — | — | — | — | — | — | 31.75 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EMAG + CADSBackbone=SD3 (MMDiT), Guidance=EMAG + CADS2025.12 | 19.95 | — | — | — | — | — | 61.9 | 27 | — | — | — | 28.86 | 26.27 | 0.95 | 0.577 | — | — | — | — | — | — | — | — | — | — | — | |
| SD3 + S2 + EMAGBackbone=SD3 (MMDiT), Guidance=S2 + EMAG2025.12 | 20.12 | — | — | — | — | — | 63.6 | 24.8 | — | — | — | 29.51 | 27.16 | 1.006 | 0.583 | — | — | — | — | — | — | — | — | — | — | — | |
| Flash SDXLEMPURPLE=No, Cosine similarity=–, Resolution=1024 x 10242026.07 | 20.42 | — | — | — | — | — | — | — | — | — | — | — | 31.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| S2Backbone=SD3 (MMDiT), Guidance=S22025.12 | 20.82 | — | — | — | — | — | 69.5 | 27.3 | — | — | — | 29.15 | 27.52 | 1.21 | 0.617 | — | — | — | — | — | — | — | — | — | — | — | |
| Hyper SDXLEMPURPLE=Yes, Cosine similarity=No, Resolution=1024 x 10242026.07 | 20.85 | — | — | — | — | — | — | — | — | — | — | — | 30.04 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SDXL-turboEMPURPLE=Yes, Cosine similarity=No, Resolution=512 x 5122026.07 | 21.24 | — | — | — | — | — | — | — | — | — | — | — | 31.54 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| APGBackbone=SD3 (MMDiT), Guidance=APG2025.12 | 21.82 | — | — | — | — | — | 70.3 | 29.5 | — | — | — | 29.45 | 26.59 | 1.235 | 0.624 | — | — | — | — | — | — | — | — | — | — | — | |
| EMAG + APGBackbone=SD3 (MMDiT), Guidance=EMAG + APG2025.12 | 21.82 | — | — | — | — | — | 68.5 | 25.8 | — | — | — | 29.79 | 26.55 | 1.174 | 0.613 | — | — | — | — | — | — | — | — | — | — | — | |
| Hyper SDXLEMPURPLE=Yes, Cosine similarity=Yes, Resolution=1024 x 10242026.07 | 21.95 | — | — | — | — | — | — | — | — | — | — | — | 30.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EMAG-IBackbone=SD3 (MMDiT), Guidance=EMAG-I2025.12 | 22.15 | — | — | — | — | — | 66.3 | 26.3 | — | — | — | 29.56 | 26.63 | 1.082 | 0.595 | — | — | — | — | — | — | — | — | — | — | — | |
| DPMSolver++Steps=42026.06 | 22.36 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DPMSolverSteps=82026.06 | 22.44 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SDXL-LightningEMPURPLE=No, Cosine similarity=–, Resolution=1024 x 10242026.07 | 22.53 | — | — | — | — | — | — | — | — | — | — | — | 31.25 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CFGBackbone=SD3 (MMDiT), Guidance=CFG2025.12 | 22.88 | — | — | — | — | — | 69.4 | 27.7 | — | — | — | 29.22 | 26.61 | 1.176 | 0.615 | — | — | — | — | — | — | — | — | — | — | — | |
| EMAGBackbone=SD3 (MMDiT), Guidance=EMAG2025.12 | 22.98 | — | — | — | — | — | 64.4 | 23.9 | — | — | — | 29.68 | 26.72 | 1.021 | 0.568 | — | — | — | — | — | — | — | — | — | — | — | |
| SEG + CFGBackbone=SD3 (MMDiT), Guidance=SEG + CFG2025.12 | 23.45 | — | — | — | — | — | 65.6 | 20.6 | — | — | — | 29.25 | 25.45 | 1.012 | 0.578 | — | — | — | — | — | — | — | — | — | — | — | |
| LCM-LoRASteps=42026.06 | 23.62 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| No CFGBackbone=SD3 (MMDiT), Guidance=None2025.12 | 23.86 | — | — | — | — | — | 49.4 | 37.1 | — | — | — | 21.61 | 23.79 | 0.591 | 0.505 | — | — | — | — | — | — | — | — | — | — | — | |
| ERGBackbone=SD3 (MMDiT), Guidance=ERG2025.12 | 23.99 | — | — | — | — | — | 66 | 20.9 | — | — | — | 29.35 | 25.71 | 1.025 | 0.58 | — | — | — | — | — | — | — | — | — | — | — | |
| Hyper SDXLEMPURPLE=No, Cosine similarity=–, Resolution=1024 x 10242026.07 | 24.97 | — | — | — | — | — | — | — | — | — | — | — | 31.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SDXL-turboEMPURPLE=No, Cosine similarity=–, Resolution=512 x 5122026.07 | 25.21 | — | — | — | — | — | — | — | — | — | — | — | 31.71 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SD3-MediumNFE=50 × 22026.07 | 28.26 | — | — | — | — | — | — | — | — | — | — | — | 33.5384 | — | — | — | — | — | — | — | — | — | — | — | — | 10.9 | |
| LCMNFE=42026.07 | 30.59 | — | — | — | — | — | — | — | — | — | — | — | 30.62 | — | — | — | — | — | — | — | — | — | — | — | — | 6.68 | |
| PFMNFE=42026.07 | 31.16 | — | — | — | — | — | — | — | — | — | — | — | 31.89 | — | — | — | — | — | — | — | — | — | — | — | — | 8.5 | |
| RASBackbone=Stable Diffusion 3, Steps=7, Sample Ratio=25.0%, Image/s=1.452025.02 | 31.99 | — | — | — | — | — | — | — | — | — | — | — | 30.64 | — | — | — | — | 21.7 | — | — | — | — | — | — | — | — | |
| RASBackbone=Stable Diffusion 3, Steps=7, Sample Ratio=12.5%, Image/s=1.482025.02 | 32.86 | — | — | — | — | — | — | — | — | — | — | — | 30.55 | — | — | — | — | 22.1 | — | — | — | — | — | — | — | — | |
| RASBackbone=Stable Diffusion 3, Steps=6, Sample Ratio=25.0%, Image/s=1.522025.02 | 33.24 | — | — | — | — | — | — | — | — | — | — | — | 30.38 | — | — | — | — | 21.51 | — | — | — | — | — | — | — | — | |
| RASBackbone=Stable Diffusion 3, Steps=6, Sample Ratio=12.5%, Image/s=1.572025.02 | 33.81 | — | — | — | — | — | — | — | — | — | — | — | 30.33 | — | — | — | — | 21.62 | — | — | — | — | — | — | — | — | |
| PFMNFE=82026.07 | 33.93 | — | — | — | — | — | — | — | — | — | — | — | 31.7 | — | — | — | — | — | — | — | — | — | — | — | — | 11.42 | |
| LCMNFE=82026.07 | 34.18 | — | — | — | — | — | — | — | — | — | — | — | 30.42 | — | — | — | — | — | — | — | — | — | — | — | — | 9.19 | |
| Q-Diffusion + ComposerBitwidth (W/A)=2/82026.03 | 35.26 | — | — | — | — | — | 55.2 | — | — | — | — | — | — | — | — | — | 78.21 | 14.5 | — | — | — | — | — | — | — | — | |
| DMD2NFE=82026.07 | 36.16 | — | — | — | — | — | — | — | — | — | — | — | 31.54 | — | — | — | — | — | — | — | — | — | — | — | — | 10.19 | |
| DMD2NFE=42026.07 | 36.2 | — | — | — | — | — | — | — | — | — | — | — | 31.24 | — | — | — | — | — | — | — | — | — | — | — | — | 8.28 | |
| RFlowBackbone=Stable Diffusion 3, Steps=5, Sample Ratio=100%, Image/s=1.432025.02 | 39.7 | — | — | — | — | — | — | — | — | — | — | — | 29.84 | — | — | — | — | 22.34 | — | — | — | — | — | — | — | — | |
| Q-DiffusionBitwidth (W/A)=2/82026.03 | 43.36 | — | — | — | — | — | 43.18 | — | — | — | — | — | — | — | — | — | 49.08 | 17.15 | — | — | — | — | — | — | — | — | |
| RASBackbone=Lumina-Next-T2I, Steps=10, Sample Ratio=25.0%, Image/s=0.592025.02 | 45.67 | — | — | — | — | — | — | — | — | — | — | — | 29.82 | — | — | — | — | 32.36 | — | — | — | — | — | — | — | — | |
| RASBackbone=Lumina-Next-T2I, Steps=10, Sample Ratio=12.5%, Image/s=0.652025.02 | 47.34 | — | — | — | — | — | — | — | — | — | — | — | 29.75 | — | — | — | — | 32.69 | — | — | — | — | — | — | — | — | |
| RFlowBackbone=Lumina-Next-T2I, Steps=7, Sample Ratio=100%, Image/s=0.492025.02 | 48.19 | — | — | — | — | — | — | — | — | — | — | — | 28.65 | — | — | — | — | 38.6 | — | — | — | — | — | — | — | — | |
| RASBackbone=Stable Diffusion 3, Steps=5, Sample Ratio=25.0%, Image/s=1.942025.02 | 51.92 | — | — | — | — | — | — | — | — | — | — | — | 29.06 | — | — | — | — | 25.67 | — | — | — | — | — | — | — | — | |
| RASBackbone=Stable Diffusion 3, Steps=5, Sample Ratio=12.5%, Image/s=1.992025.02 | 53.24 | — | — | — | — | — | — | — | — | — | — | — | 28.94 | — | — | — | — | 26.04 | — | — | — | — | — | — | — | — | |
| RASBackbone=Lumina-Next-T2I, Steps=7, Sample Ratio=25.0%, Image/s=0.702025.02 | 53.93 | — | — | — | — | — | — | — | — | — | — | — | 28.85 | — | — | — | — | 39.8 | — | — | — | — | — | — | — | — | |
| RASBackbone=Lumina-Next-T2I, Steps=7, Sample Ratio=12.5%, Image/s=0.742025.02 | 54.62 | — | — | — | — | — | — | — | — | — | — | — | 28.83 | — | — | — | — | 40.23 | — | — | — | — | — | — | — | — | |
| RFlowBackbone=Stable Diffusion 3, Steps=4, Sample Ratio=100%, Image/s=1.792025.02 | 61.92 | — | — | — | — | — | — | — | — | — | — | — | 28.45 | — | — | — | — | 27.42 | — | — | — | — | — | — | — | — | |
| RASBackbone=Lumina-Next-T2I, Steps=6, Sample Ratio=25.0%, Image/s=0.752025.02 | 67.16 | — | — | — | — | — | — | — | — | — | — | — | 27.85 | — | — | — | — | 46.46 | — | — | — | — | — | — | — | — | |
| RASBackbone=Lumina-Next-T2I, Steps=6, Sample Ratio=12.5%, Image/s=0.782025.02 | 67.88 | — | — | — | — | — | — | — | — | — | — | — | 27.83 | — | — | — | — | 45.88 | — | — | — | — | — | — | — | — | |
| RFlowBackbone=Lumina-Next-T2I, Steps=5, Sample Ratio=100.%, Image/s=0.692025.02 | 96.53 | — | — | — | — | — | — | — | — | — | — | — | 26.03 | — | — | — | — | 59.26 | — | — | — | — | — | — | — | — | |
| (No CFG)Guidance=None, Backbone=SD32025.12 | — | 23.859 | — | — | — | — | — | — | — | — | — | 21.61 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ADD (SD-Turbo)Teacher=SD2.12024.06 | — | — | 27.48 | 26.89 | 26.86 | 27.46 | 65 | 35 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AMDbase model=SDXL2026.02 | — | — | — | — | — | — | — | — | 88.37 | 31.25 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| APGGuidance=Adaptive Projective Guidance, Backbone=SD32025.12 | — | 21.816 | — | — | — | — | — | — | — | — | — | 29.45 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Base Model (SD3.5-Large)2026.03 | — | — | — | — | — | — | — | — | 1.051 | — | — | — | — | — | — | — | — | — | 1.819 | 90.5 | — | — | — | — | — | — | |
| BOOTTeacher=SD1.52024.06 | — | — | 25.29 | 25.16 | 24.4 | 24.61 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CADSGuidance=Context-Aware Diffusion Sampling, Backbone=SD32025.12 | — | 18.32 | — | — | — | — | — | — | — | — | — | 28.36 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CADSs=10^-122026.03 | — | — | — | — | — | — | — | — | 0.131 | — | — | — | — | — | — | — | — | — | 2.004 | 71.7 | 0.941 | — | — | — | — | — | |
| CADSs=10^-102026.03 | — | — | — | — | — | — | — | — | 0.051 | — | — | — | — | — | — | — | — | — | 2.025 | 69.2 | 0.953 | — | — | — | — | — | |
| CADSs=10^-082026.03 | — | — | — | — | — | — | — | — | 0.066 | — | — | — | — | — | — | — | — | — | 2.018 | 69.2 | 0.953 | — | — | — | — | — | |
| CaT-S0#MACs=3G, #Params=169M2024.04 | — | 5.49 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CaT-S1#MACs=7G, #Params=307M2024.04 | — | 5.22 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CFGBackbone=SD3, Sampling steps=Identical2025.12 | — | — | — | — | — | — | — | — | — | 29.22 | 29.68 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CFGGuidance=Classifier-Free Guidance, Backbone=SD32025.12 | — | 22.877 | — | — | — | — | — | — | — | — | — | 29.22 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Contextual Space repulsioneta=2.5 · 10^42026.03 | — | — | — | — | — | — | — | — | 1.018 | — | — | — | — | — | — | — | — | — | 1.851 | 90.4 | 0.619 | — | — | — | — | — | |
| Contextual Space repulsioneta=2.5 · 10^62026.03 | — | — | — | — | — | — | — | — | 1.012 | — | — | — | — | — | — | — | — | — | 1.878 | 90.4 | 0.627 | — | — | — | — | — | |
| Contextual Space repulsioneta=2.5 · 10^72026.03 | — | — | — | — | — | — | — | — | 0.988 | — | — | — | — | — | — | — | — | — | 1.941 | 90 | 0.625 | — | — | — | — | — | |
| Contextual Space repulsioneta=2.5 · 10^82026.03 | — | — | — | — | — | — | — | — | 0.94 | — | — | — | — | — | — | — | — | — | 1.98 | 89 | 0.445 | — | — | — | — | — | |
| D-DMDbase model=SDXL2026.02 | — | — | — | — | — | — | — | — | 78.61 | 30.34 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DMD2base model=SDXL2026.02 | — | — | — | — | — | — | — | — | 71.01 | 30.64 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EMAGBackbone=SD3, Sampling steps=Identical2025.12 | — | — | — | — | — | — | — | — | — | 29.68 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EMAGGuidance=EMAG, Backbone=SD32025.12 | — | 22.98 | — | — | — | — | — | — | — | — | — | 29.68 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EMAG + APGGuidance=EMAG + APG, Backbone=SD32025.12 | — | 21.819 | — | — | — | — | — | — | — | — | — | 29.79 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EMAG + CADSGuidance=EMAG + CADS, Backbone=SD32025.12 | — | 19.95 | — | — | — | — | — | — | — | — | — | 28.86 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EMAG-IGuidance=EMAG-I, Backbone=SD32025.12 | — | 22.154 | — | — | — | — | — | — | — | — | — | 29.56 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ERGBackbone=SD3, Sampling steps=Identical2025.12 | — | — | — | — | — | — | — | — | — | 29.35 | 29.56 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ERGGuidance=ERG, Backbone=SD32025.12 | — | 23.989 | — | — | — | — | — | — | — | — | — | 29.35 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Flashbase model=SDXL2026.02 | — | — | — | — | — | — | — | — | 19.04 | 27.71 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Friro#Params=512M2024.04 | — | 8.97 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GoT-R1-7Btraining_stage=GRPO-enhanced2025.05 | — | — | — | — | — | — | — | — | — | — | — | — | 31.83 | — | — | — | — | — | — | — | — | 5.41 | 77 | — | — | — | |
| InstaFlowTeacher=SD1.52024.06 | — | — | 25.98 | 26.32 | 25.79 | 25.93 | 53 | 45 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — |