Image Generation on ImageNet 256x256 (val)
0.58FIDVQGAN-f-4
Evaluation Results
| Method | Links | |||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| VQGAN-f-4mode=first stage, Generator Compute (V100 days)=29, Overall Compute (V100 days)=29, Nparams=55M2021.12 | 0.58 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| JiT-H + FD-lossNFE=1, space=pixel, #params=953M2026.04 | 0.75 | 313 | — | 76 | 66 | — | — | — | — | — | — | — | 2.65 | — | — | — | — | |
| iMF-XL + FD-lossNFE=1, space=latent, #params=610M2026.04 | 0.76 | 301.3 | — | 77 | 67 | — | — | — | — | — | — | — | 2.45 | — | — | — | — | |
| JiT-L + FD-lossNFE=1, space=pixel, #params=459M2026.04 | 0.77 | 317.3 | — | 77 | 66 | — | — | — | — | — | — | — | 3.24 | — | — | — | — | |
| pMF-H + FD-lossNFE=1, space=pixel, #params=935M2026.04 | 0.77 | 310.1 | — | 77 | 68 | — | — | — | — | — | — | — | 1.89 | — | — | — | — | |
| pMF-L + FD-lossNFE=1, space=pixel, #params=410M2026.04 | 0.78 | 309.2 | — | 76 | 67 | — | — | — | — | — | — | — | 2.09 | — | — | — | — | |
| BAR-LNFE=256×2×4, space=discrete, #params=1.1B2026.04 | 1.01 | 281.9 | — | 77 | 68 | — | — | — | — | — | — | — | 3.57 | — | — | — | — | |
| REPA-EEpochs=800, Backbone=SiT-XL, Sampling protocol=Class-balanced, Fine-tuning autoencoders=true, Guidance Scale (w)=1.82026.02 | 1.12 | 302.9 | — | 79 | 66 | 4.09 | — | — | — | — | — | — | — | — | — | — | — | |
| VQGAN-f-8mode=first stage, Generator Compute (V100 days)=66, Overall Compute (V100 days)=66, Nparams=68M2021.12 | 1.14 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RAE-XLNFE=50×2†, space=latent, #params=839M2026.04 | 1.16 | 261 | — | 77 | 67 | — | — | — | — | — | — | — | 3.26 | — | — | — | — | |
| REPA-ENFE=250×2†, space=latent, #params=676M2026.04 | 1.17 | 298.3 | — | 79 | 66 | — | — | — | — | — | — | — | 3.04 | — | — | — | — | |
| xAR-Htype=xAR, Parameters=1.1B, Sampling Steps=502025.02 | 1.24 | — | — | — | — | — | — | — | — | — | 1.3 | — | — | — | — | — | — | |
| xAR-Htype=xAR, #params=1.1B2025.02 | 1.24 | 301.6 | — | 83 | 64 | — | — | — | — | — | — | — | — | — | — | — | — | |
| DDT-XLNFE=250×2, space=latent, #params=675M2026.04 | 1.26 | 309.3 | — | 79 | 66 | — | — | — | — | — | — | — | 5.7 | — | — | — | — | |
| xAR-Ltype=xAR, Parameters=608M, Sampling Steps=502025.02 | 1.28 | — | — | — | — | — | — | — | — | — | 3.2 | — | — | — | — | — | — | |
| xAR-Ltype=xAR, #params=608M2025.02 | 1.28 | 292.5 | — | 82 | 62 | — | — | — | — | — | — | — | — | — | — | — | — | |
| Ours (StableVM/VA-REPA)Epochs=400, Backbone=SiT-XL, Sampling protocol=Class-balanced, Guidance Scale (w)=1.82026.02 | 1.34 | 305 | — | 80 | 64 | 4.53 | — | — | — | — | — | — | — | — | — | — | — | |
| SiD2, Flop HeavyNFE=5122024.10 | 1.38 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SiD2NFE=512×2, Space=pixel2026.06 | 1.38 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAR-L, DeTokNFE=256×2×100, space=latent, #params=478M2026.04 | 1.39 | 306.2 | — | 81 | 62 | — | — | — | — | — | — | — | 5.49 | — | — | — | — | |
| REGEpochs=480, Backbone=SiT-XL, Guidance Scale (w)=1.82026.02 | 1.4 | 296.9 | — | 77 | 66 | 4.24 | — | — | — | — | — | — | — | — | — | — | — | |
| REPAtype=Flow., Parameters=675M, Sampling Steps=2502025.02 | 1.42 | — | — | — | — | — | — | — | — | — | 0.6 | — | — | — | — | — | — | |
| REPAtype=Flow-Matching, #params=675M, guidance interval sampling=true2025.02 | 1.42 | 305.7 | — | 80 | 65 | — | — | — | — | — | — | — | — | — | — | — | — | |
| REPAEpochs=800, Backbone=SiT-XL, Guidance Scale (w)=1.82026.02 | 1.42 | 305.7 | — | 80 | 65 | 4.7 | — | — | — | — | — | — | — | — | — | — | — | |
| REPA-XL/2Params=675M + 86M, Epochs=800, NFE=250x2, Latency(s)=3.44, Type=Latent Diffusion Model (VAE)2025.11 | 1.42 | 305.7 | — | 80 | 64 | 4.7 | — | — | — | — | — | — | — | — | — | — | — | |
| SiT-XL/2-REPANFE=250×2†, space=latent, #params=675M2026.04 | 1.42 | 306.1 | — | 80 | 65 | — | — | — | — | — | — | — | 5.45 | — | — | — | — | |
| LightningDiTNFE=250×2, space=latent, #params=675M2026.04 | 1.42 | 294.3 | — | 80 | 64 | — | — | — | — | — | — | — | 4.57 | — | — | — | — | |
| REPA-XL/2Params=675M + 86M, Epochs=800, GFLOPs=2382026.05 | 1.42 | 305.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SiT-XL/2 + REPANFE=250×2, Space=latent, Params=675M2026.06 | 1.42 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Drift-L (pixel)NFE=1, space=pixel, #params=465M2026.04 | 1.43 | 305.8 | — | 81 | 60 | — | — | — | — | — | — | — | 10.51 | — | — | — | — | |
| Ours (StableVM/VA-REPA)Epochs=400, Backbone=SiT-XL, Sampling protocol=Standard REPA, Guidance Scale (w)=1.82026.02 | 1.47 | 300.3 | — | 80 | 63 | 4.51 | — | — | — | — | — | — | — | — | — | — | — | |
| RAR-XXLtype=AR, #params=1.5B2025.02 | 1.48 | 326 | — | 80 | 63 | — | — | — | — | — | — | — | — | — | — | — | — | |
| MaskBittype=Mask., #params=305M2025.02 | 1.52 | 328.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Drift-L (latent)NFE=1, space=latent, #params=463M2026.04 | 1.53 | 257.2 | — | 79 | 63 | — | — | — | — | — | — | — | 10.92 | — | — | — | — | |
| REGNFE=250×2†, space=latent, #params=685M2026.04 | 1.54 | 302.9 | — | 78 | 62 | — | — | — | — | — | — | — | 4.64 | — | — | — | — | |
| MAR-Htype=MAR, Parameters=943M, Sampling Steps=2562025.02 | 1.55 | — | — | — | — | — | — | — | — | — | 0.3 | — | — | — | — | — | — | |
| MAR-Htype=MAR, #params=943M2025.02 | 1.55 | 303.7 | — | 81 | 62 | — | — | — | — | — | — | — | — | — | — | — | — | |
| PaGoDANFE=12024.10 | 1.56 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAR-HNFE=256×2×100, space=latent, #params=942M2026.04 | 1.56 | 299.5 | — | 80 | 62 | — | — | — | — | — | — | — | 5.61 | — | — | — | — | |
| CausalFusion-H†type=Diffusion, #params=1B, guidance interval sampling=true2025.02 | 1.57 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MDTv2-XL/2type=Diffusion, #params=676M2025.02 | 1.58 | 314.7 | — | 79 | 65 | — | — | — | — | — | — | — | — | — | — | — | — | |
| SiT-XL/2 + routing + SGFID#Epoch=4002026.01 | 1.58 | 249.7 | — | 80 | 63 | 4.45 | — | — | — | — | — | — | — | — | — | — | — | |
| iMF-XLNFE=2, space=latent, #params=610M2026.04 | 1.61 | 289.1 | — | 79 | 63 | — | — | — | — | — | — | — | 7.48 | — | — | — | — | |
| PixelDiT-XLParams=797M, Epochs=320, GFLOPs=3112026.05 | 1.61 | 292.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PixelDiT-XL/16NFE=100×2, Space=pixel, Params=797M2026.06 | 1.61 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeCo-XL/16Params=682M, Epochs=800, NFE=250x2, Latency(s)=2.632025.11 | 1.62 | 301 | — | 80 | 62 | 4.41 | — | — | — | — | — | — | — | — | — | — | — | |
| CrossFlow-XLNFE=1, Space=latent→pixel, Params=461M2026.06 | 1.62 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DiMR-G/2Rtype=Diffusion, #params=1.1B2025.02 | 1.63 | 292.5 | — | 79 | 63 | — | — | — | — | — | — | — | — | — | — | — | — | |
| FlowAR-Htype=AR, #params=1.9B2025.02 | 1.65 | 296.5 | — | 83 | 60 | — | — | — | — | — | — | — | — | — | — | — | — | |
| REPA-EEpochs=80, Backbone=SiT-XL, Sampling protocol=Class-balanced, Fine-tuning autoencoders=true, Guidance Scale (w)=1.82026.02 | 1.67 | 266.3 | — | 80 | 63 | 4.12 | — | — | — | — | — | — | — | — | — | — | — | |
| reference (real images)NFE=50k validation images, space=N/A, #params=N/A2026.04 | 1.68 | 232.2 | — | 75 | 66 | — | — | — | — | — | — | — | 1 | — | — | — | — | |
| FlowAR-HNFE=50×2†, space=latent, #params=1.9B2026.04 | 1.68 | 274.1 | — | 80 | 62 | — | — | — | — | — | — | — | 6.13 | — | — | — | — | |
| SiT-XL/2+REPA-EParam=675M, Epoch=800, Guidance=None, NFE=250, latent_space=representational2025.11 | 1.69 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BitDance-B-4xType=AR, Order=raster, #Params=260M, Steps=64, Throughput=24.18 img/s, Parallel tokens per step=42026.02 | 1.69 | 291.2 | — | 79 | 63 | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeCo-XL/16 (Heun)Params=682M, Epochs=600, NFE=100x2, Latency(s)=1.05, Sampler=Heun2025.11 | 1.69 | 304 | — | 79 | 63 | 4.59 | — | — | — | — | — | — | — | — | — | — | — | |
| CrossFlow-LNFE=1, Space=latent→pixel, Params=315M2026.06 | 1.69 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Ours (StableVM/VA-REPA)Epochs=80, Backbone=SiT-XL, Sampling protocol=Class-balanced, Guidance Scale (w)=1.82026.02 | 1.71 | 274.2 | — | 81 | 61 | 4.54 | — | — | — | — | — | — | — | — | — | — | — | |
| DeCo-XL/16Params=682M, Epochs=800, NFE=100x2, Latency(s)=1.052025.11 | 1.71 | 304 | — | 80 | 61 | 4.54 | — | — | — | — | — | — | — | — | — | — | — | |
| xAR-Btype=xAR, Parameters=172M, Sampling Steps=502025.02 | 1.72 | — | — | — | — | — | — | — | — | — | 9.8 | — | — | — | — | — | — | |
| xAR-Btype=xAR, #params=172M2025.02 | 1.72 | 280.4 | — | 82 | 59 | — | — | — | — | — | — | — | — | — | — | — | — | |
| SiD2, smallNFE=5122024.10 | 1.72 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| iMF-XL/2NFE=1, Space=latent, Params=659M2026.06 | 1.72 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAR-Ltype=AR, Parameters=479M, Sampling Steps=2562025.02 | 1.78 | — | — | — | — | — | — | — | — | — | 0.5 | — | — | — | — | — | — | |
| MAGVIT-v2type=Mask., #params=307M2025.02 | 1.78 | 319.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAR-Ltype=MAR, #params=479M2025.02 | 1.78 | 296 | — | 81 | 60 | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAR-LType=Mask., Order=random, #Params=479M, Steps=256, Throughput=1.39 img/s2026.02 | 1.78 | 296 | — | 81 | 60 | — | — | — | — | — | — | — | — | — | — | — | — | |
| MDT-XL/2#Epoch=13002026.01 | 1.79 | 283.01 | — | 81 | 61 | 4.57 | — | — | — | — | — | — | — | — | — | — | — | |
| REPAtype=Flow-Matching, #params=675M, guidance interval sampling=false2025.02 | 1.8 | 284 | — | 81 | 61 | — | — | — | — | — | — | — | — | — | — | — | — | |
| Ours (StableVM/VA-REPA)Epochs=80, Backbone=SiT-XL, Sampling protocol=Standard REPA, Guidance Scale (w)=1.82026.02 | 1.8 | 272.4 | — | 81 | 60 | 4.52 | — | — | — | — | — | — | — | — | — | — | — | |
| SiT-XL/2 + REPA#Epoch=8002026.01 | 1.8 | 284 | — | 81 | 61 | 4.5 | — | — | — | — | — | — | — | — | — | — | — | |
| MAR-LNFE=256×2×100, space=latent, #params=478M2026.04 | 1.8 | 293.4 | — | 80 | 60 | — | — | — | — | — | — | — | 6.68 | — | — | — | — | |
| iMF-XLNFE=1, space=latent, #params=610M2026.04 | 1.82 | 278.9 | — | 78 | 63 | — | — | — | — | — | — | — | 8.39 | — | — | — | — | |
| JiT-G/16Params=2B, Epochs=600, GFLOPs=7662026.05 | 1.82 | 292.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DiT-XL/2-G++Diffusion Space=Latent2022.11 | 1.83 | 281.53 | — | 78 | 64 | 5.16 | — | 70 | — | — | — | — | — | — | — | — | — | |
| REGEpochs=80, Backbone=SiT-XL, Guidance Scale (w)=1.82026.02 | 1.86 | 321.4 | — | 76 | 63 | 4.49 | — | — | — | — | — | — | — | — | — | — | — | |
| JiT-H/16params=953M2025.12 | 1.86 | 303.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| JiT-H/16 (Heun)Params=953M, Epochs=600, NFE=100x2, Sampler=Heun2025.11 | 1.86 | 303 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RAE-XLParam=676M, Epoch=800, Guidance=None, NFE=250, latent_space=representational2025.11 | 1.87 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RDMNFE=3562024.10 | 1.89 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeCo-XL/16Params=682M, Epochs=320, NFE=100x2, Latency(s)=1.052025.11 | 1.9 | 303 | — | 80 | 61 | 4.47 | — | — | — | — | — | — | — | — | — | — | — | |
| DeCo-XL/16Params=682M, Epochs=320, GFLOPs=2372026.05 | 1.9 | 303 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BitDance-B-16xType=AR, Order=raster, #Params=260M, Steps=16, Throughput=90.26 img/s, Parallel tokens per step=162026.02 | 1.91 | 283.8 | — | 78 | 62 | — | — | — | — | — | — | — | — | — | — | — | — | |
| FREPix-XLParams=674M, Epochs=320, GFLOPs=2302026.05 | 1.91 | 295.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| iREPAEpochs=80, Backbone=SiT-XL, Guidance Scale (w)=1.82026.02 | 1.93 | 268.8 | — | 80 | 60 | 4.59 | — | — | — | — | — | — | — | — | — | — | — | |
| PixNerd-XL/16Params=700M, Epochs=320, NFE=100x2, Latency(s)=1.182025.11 | 1.95 | 300 | — | 80 | 60 | 4.54 | — | — | — | — | — | — | — | — | — | — | — | |
| TiTok-S-128type=Mask., #params=287M2025.02 | 1.97 | 281.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAR-d30type=VAR, #params=2.0B2025.02 | 1.97 | 323.1 | — | 82 | 59 | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAR-d30NFE=10×2, space=discrete, #params=2B2026.04 | 1.97 | 304.6 | — | 82 | 59 | — | — | — | — | — | — | — | 6.7 | — | — | — | — | |
| JiT-HNFE=50×2×2†, space=pixel, #params=953M2026.04 | 1.97 | 296 | — | 78 | 63 | — | — | — | — | — | — | — | 7.66 | — | — | — | — | |
| REPAEpochs=80, Backbone=SiT-XL, Guidance Scale (w)=1.82026.02 | 1.98 | 263 | — | 80 | 61 | 4.6 | — | — | — | — | — | — | — | — | — | — | — | |
| PixelFlow, XL/4params=677M2025.12 | 1.98 | 282.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PixelFlow-XL/4Params=677M, Epochs=320, NFE=120x2, Latency(s)=9.782025.11 | 1.98 | 282.1 | — | 81 | 60 | 5.83 | — | — | — | — | — | — | — | — | — | — | — | |
| PixelFlow-XL/4Params=677M, Epochs=320, GFLOPs=58182026.05 | 1.98 | 282.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RDMParams=553M + 553M, Epochs=400, NFE=250, Latency(s)=38.4, Type=Latent Diffusion Model (VAE)2025.11 | 1.99 | 260.4 | — | 81 | 58 | 3.99 | — | — | — | — | — | — | — | — | — | — | — | |
| SiT-XL/2type=Flow., Parameters=675M, Sampling Steps=2502025.02 | 2.02 | — | — | — | — | — | — | — | — | — | 0.5 | — | — | — | — | — | — | |
| Faster-DiTEpochs=400, Guidance Scale (w)=1.82026.02 | 2.03 | 264 | — | 81 | 60 | 4.63 | — | — | — | — | — | — | — | — | — | — | — | |
| FasterDiT-XL/2#Epoch=4002026.01 | 2.03 | 264 | — | 81 | 60 | 4.63 | — | — | — | — | — | — | — | — | — | — | — | |
| OUSACSteps=50, CFG-Steps=8, MACs (T)=5.07, Latency (s)=2.812025.12 | 2.04 | 266.5 | — | 82.09 | 58.7 | 4.29 | — | — | — | — | — | — | — | — | — | — | — | |
| DiCo-XLType=Diff., Order=-, #Params=701M, Steps=250, Throughput=2.62 img/s2026.02 | 2.05 | 282.2 | — | 83 | 59 | — | — | — | — | — | — | — | — | — | — | — | — | |
| SiT-XL/2type=Flow-Matching, #params=675M2025.02 | 2.06 | 277.5 | — | 83 | 59 | — | — | — | — | — | — | — | — | — | — | — | — |