Class-conditional Image Generation on ImageNet-1K 256x256 (train)
365.4ISVAR
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| VARTraining Epochs=350, Backbone Params (M)=20002025.12 | 365.4 | 1.8 | — | 83 | 57 | — | |
| FractalMAR-HTraining Epochs=600, Backbone Params (M)=8482025.12 | 348.9 | 6.15 | — | 81 | 46 | — | |
| FractalMAR-HParadigm=Autoregressive, Space=Pixel-Space, Params=848M2026.06 | 348.9 | — | — | — | — | 6.15 | |
| FractalMAR-LParadigm=Autoregressive, Space=Pixel-Space, Params=438M2026.06 | 334.9 | — | — | — | — | 7.3 | |
| MagViT-v2Training Epochs=1080, Backbone Params (M)=3072025.12 | 319.4 | 1.78 | — | — | — | — | |
| MDTv2Training Epochs=1080, Backbone Params (M)=675, VAE Params Enc. (M)=34, VAE Params Enctr (M)=34, VAE Params Dec. (M)=50, VAE Params Totaltr (M)=84, VAE Params Total (M)=842025.12 | 314.7 | 1.58 | 4.52 | 79 | 65 | — | |
| DDT-XL/2Paradigm=Diffusion, Space=Two-Stage, Params=675+49M2026.06 | 310.6 | — | — | — | — | 1.26 | |
| RAE (DiT-XL)Training Epochs=800, Backbone Params (M)=676, VAE Params Enc. (M)=87, VAE Params Enctr (M)=0, VAE Params Dec. (M)=415, VAE Params Totaltr (M)=415, VAE Params Total (M)=5022025.12 | 309.4 | 1.41 | — | 80 | 63 | — | |
| REPATraining Epochs=800, Backbone Params (M)=675, VAE Params Enc. (M)=34, VAE Params Enctr (M)=34, VAE Params Dec. (M)=50, VAE Params Totaltr (M)=84, VAE Params Total (M)=842025.12 | 305.7 | 1.42 | 4.7 | 80 | 65 | — | |
| REPA-XL/2Paradigm=Diffusion, Space=Two-Stage, Params=675+49M2026.06 | 305.7 | — | — | — | — | 1.42 | |
| DeCoTraining Epochs=600, Backbone Params (M)=6822025.12 | 304 | 1.69 | 4.59 | 79 | 63 | — | |
| MARTraining Epochs=800, Backbone Params (M)=9452025.12 | 303.7 | 1.55 | — | 81 | 62 | — | |
| JiTTraining Epochs=600, Backbone Params (M)=9532025.12 | 303 | 1.86 | — | — | — | — | |
| DeCoTraining Epochs=320, Backbone Params (M)=6822025.12 | 303 | 1.9 | 4.47 | 80 | 61 | — | |
| VAR-d20Paradigm=Autoregressive, Space=Two-Stage, Params=600M+40M2026.06 | 302.6 | — | — | — | — | 2.57 | |
| PixNerdTraining Epochs=320, Backbone Params (M)=7002025.12 | 300 | 1.95 | 4.54 | 80 | 60 | — | |
| JiT-L/16Paradigm=Diffusion, Space=Pixel-Space, Params=459M2026.06 | 298.5 | — | — | — | — | 2.36 | |
| PixNerd-L/16Paradigm=Diffusion, Space=Pixel-Space, Params=458M2026.06 | 297 | — | — | — | — | 2.64 | |
| MAR-LParadigm=Autoregressive, Space=Two-Stage, Params=479+49M2026.06 | 296 | — | — | — | — | 1.78 | |
| SphereAR-LParadigm=Autoregressive, Space=Two-Stage, Params=479+49M2026.06 | 295.9 | — | — | — | — | 1.54 | |
| VA-VAETraining Epochs=800, Backbone Params (M)=675, VAE Params Enc. (M)=28, VAE Params Enctr (M)=28, VAE Params Dec. (M)=41, VAE Params Totaltr (M)=70, VAE Params Total (M)=702025.12 | 295.3 | 1.35 | 4.15 | 79 | 65 | — | |
| LightningDiT-XL/2Paradigm=Diffusion, Space=Two-Stage, Params=675+49M2026.06 | 295.3 | — | — | — | — | 1.35 | |
| PixelDiTTraining Epochs=320, Backbone Params (M)=7972025.12 | 292.7 | 1.61 | 4.68 | 78 | 64 | — | |
| PRA-LParadigm=Autoregressive, Space=Pixel-Space, Params=511M2026.06 | 287.3 | — | — | — | — | 1.94 | |
| MDTTraining Epochs=1300, Backbone Params (M)=675, VAE Params Enc. (M)=34, VAE Params Enctr (M)=34, VAE Params Dec. (M)=50, VAE Params Totaltr (M)=84, VAE Params Total (M)=842025.12 | 283 | 1.79 | 4.57 | 81 | 61 | — | |
| PixelDiTTraining Epochs=80, Backbone Params (M)=7972025.12 | 282.3 | 2.36 | 5.11 | 80 | 57 | — | |
| PixelFlowTraining Epochs=320, Backbone Params (M)=6772025.12 | 282.1 | 1.98 | 5.83 | 81 | 60 | — | |
| PixelFlow-XL/4Paradigm=Diffusion, Space=Pixel-Space, Params=677M2026.06 | 282.1 | — | — | — | — | 1.98 | |
| MAR-BParadigm=Autoregressive, Space=Two-Stage, Params=208+49M2026.06 | 281.7 | — | — | — | — | 2.31 | |
| DiTTraining Epochs=1400, Backbone Params (M)=675, VAE Params Enc. (M)=34, VAE Params Enctr (M)=34, VAE Params Dec. (M)=50, VAE Params Totaltr (M)=84, VAE Params Total (M)=842025.12 | 278.2 | 2.27 | 4.6 | 83 | 57 | — | |
| DiT-XL/2Paradigm=Diffusion, Space=Two-Stage, Params=675+49M2026.06 | 278.2 | — | — | — | — | 2.27 | |
| SiT-XL/2Paradigm=Diffusion, Space=Two-Stage, Params=675+49M2026.06 | 277.5 | — | — | — | — | 2.06 | |
| PRA-BParadigm=Autoregressive, Space=Pixel-Space, Params=250M2026.06 | 276.9 | — | — | — | — | 2.21 | |
| MaskDiTTraining Epochs=1600, Backbone Params (M)=675, VAE Params Enc. (M)=34, VAE Params Enctr (M)=34, VAE Params Dec. (M)=50, VAE Params Totaltr (M)=84, VAE Params Total (M)=842025.12 | 276.6 | 2.28 | 5.67 | 80 | 61 | — | |
| JiT-B/16Paradigm=Diffusion, Space=Pixel-Space, Params=131M2026.06 | 275.1 | — | — | — | — | 3.66 | |
| FractalMAR-BParadigm=Autoregressive, Space=Pixel-Space, Params=186M2026.06 | 274.3 | — | — | — | — | 11.8 | |
| FAETraining Epochs=800, Backbone Params (M)=675, VAE Params Enc. (M)=1138, VAE Params Enctr (M)=38, VAE Params Dec. (M)=476, VAE Params Totaltr (M)=514, VAE Params Total (M)=16142025.12 | 274.1 | 1.41 | 4.34 | 81 | 63 | — | |
| PRA-SParadigm=Autoregressive, Space=Pixel-Space, Params=135M2026.06 | 273.9 | — | — | — | — | 2.58 | |
| RePackTraining Epochs=80, Backbone Params (M)=675, VAE Params Enc. (M)=85.71, VAE Params Enctr (M)=0.04, VAE Params Dec. (M)=41.42, VAE Params Totaltr (M)=41.46, VAE Params Total (M)=127.132025.12 | 273.8 | 1.77 | 4.42 | 82 | 59 | — | |
| RePack + RefinerTraining Epochs=80, Backbone Params (M)=675, VAE Params Enc. (M)=85.71, VAE Params Enctr (M)=0.04, VAE Params Dec. (M)=41.42, VAE Params Totaltr (M)=97.04, VAE Params Total (M)=182.712025.12 | 272.6 | 1.61 | 4.4 | 81 | 60 | — | |
| SiTTraining Epochs=1400, Backbone Params (M)=675, VAE Params Enc. (M)=34, VAE Params Enctr (M)=34, VAE Params Dec. (M)=50, VAE Params Totaltr (M)=84, VAE Params Total (M)=842025.12 | 270.3 | 2.06 | 4.5 | 82 | 59 | — | |
| RePackTraining Epochs=64, Backbone Params (M)=675, VAE Params Enc. (M)=85.71, VAE Params Enctr (M)=0.04, VAE Params Dec. (M)=41.42, VAE Params Totaltr (M)=41.46, VAE Params Total (M)=127.132025.12 | 269.6 | 1.82 | 4.41 | 83 | 59 | — | |
| FARMER-1.9B/8Paradigm=Autoregressive, Space=Pixel-Space, Params=1.9B2026.06 | 269.2 | — | — | — | — | 3.6 | |
| RePack + RefinerTraining Epochs=64, Backbone Params (M)=675, VAE Params Enc. (M)=85.71, VAE Params Enctr (M)=0.04, VAE Params Dec. (M)=41.42, VAE Params Totaltr (M)=97.04, VAE Params Total (M)=182.712025.12 | 268.9 | 1.65 | 4.35 | 81 | 59 | — | |
| FAE w/ Time. ShiftTraining Epochs=800, Backbone Params (M)=675, VAE Params Enc. (M)=1138, VAE Params Enctr (M)=38, VAE Params Dec. (M)=476, VAE Params Totaltr (M)=514, VAE Params Total (M)=16142025.12 | 268 | 1.29 | 4.32 | 80 | 64 | — | |
| SVG-XL*Training Epochs=1400, Backbone Params (M)=675, VAE Params Enc. (M)=40, VAE Params Enctr (M)=11, VAE Params Dec. (M)=43, VAE Params Totaltr (M)=54, VAE Params Total (M)=83, Sampling=25-step Euler2025.12 | 264.9 | 1.92 | — | — | — | — | |
| FasterDiTTraining Epochs=400, Backbone Params (M)=675, VAE Params Enc. (M)=34, VAE Params Enctr (M)=34, VAE Params Dec. (M)=50, VAE Params Totaltr (M)=84, VAE Params Total (M)=842025.12 | 264 | 2.03 | 4.63 | 81 | 60 | — | |
| LlamaGenTraining Epochs=300, Backbone Params (M)=3100, VAE Params Total (M)=722025.12 | 263.3 | 2.18 | 5.97 | 81 | 58 | — | |
| RAE (DiTDH-XL)Training Epochs=800, Backbone Params (M)=839, VAE Params Enc. (M)=87, VAE Params Enctr (M)=0, VAE Params Dec. (M)=415, VAE Params Totaltr (M)=415, VAE Params Total (M)=5022025.12 | 262.6 | 1.13 | — | 78 | 67 | — | |
| RAE-XL/2Paradigm=Diffusion, Space=Two-Stage, Params=839+415M2026.06 | 262.6 | — | — | — | — | 1.13 | |
| SVG-XL*Training Epochs=500, Backbone Params (M)=675, VAE Params Enc. (M)=40, VAE Params Enctr (M)=11, VAE Params Dec. (M)=43, VAE Params Totaltr (M)=54, VAE Params Total (M)=83, Sampling=25-step Euler2025.12 | 258.7 | 2.1 | — | — | — | — | |
| SiDParadigm=Diffusion, Space=Pixel-Space, Params=2B2026.06 | 256.3 | — | — | — | — | 2.44 | |
| LlamaGen-XLParadigm=Autoregressive, Space=Two-Stage, Params=1.4B+70M2026.06 | 253.9 | — | — | — | — | 2.34 | |
| VA-VAETraining Epochs=64, Backbone Params (M)=675, VAE Params Enc. (M)=28, VAE Params Enctr (M)=28, VAE Params Dec. (M)=41, VAE Params Totaltr (M)=70, VAE Params Total (M)=702025.12 | 252.3 | 2.11 | 4.16 | 81 | 58 | — | |
| FAETraining Epochs=64, Backbone Params (M)=675, VAE Params Enc. (M)=1138, VAE Params Enctr (M)=38, VAE Params Dec. (M)=476, VAE Params Totaltr (M)=514, VAE Params Total (M)=16142025.12 | 250.3 | 2.01 | 4.39 | 83 | 59 | — | |
| FAETraining Epochs=80, Backbone Params (M)=675, VAE Params Enc. (M)=1138, VAE Params Enctr (M)=38, VAE Params Dec. (M)=476, VAE Params Totaltr (M)=514, VAE Params Total (M)=16142025.12 | 249.6 | 1.92 | 4.35 | 83 | 59 | — | |
| FAE w/ Time. ShiftTraining Epochs=80, Backbone Params (M)=675, VAE Params Enc. (M)=1138, VAE Params Enctr (M)=38, VAE Params Dec. (M)=476, VAE Params Totaltr (M)=514, VAE Params Total (M)=16142025.12 | 243.8 | 1.7 | 4.33 | 82 | 61 | — | |
| FAE w/ Time. ShiftTraining Epochs=64, Backbone Params (M)=675, VAE Params Enc. (M)=1138, VAE Params Enctr (M)=38, VAE Params Dec. (M)=476, VAE Params Totaltr (M)=514, VAE Params Total (M)=16142025.12 | 241.1 | 1.87 | 4.39 | 82 | 59 | — | |
| FARMER-1.1B/8Paradigm=Autoregressive, Space=Pixel-Space, Params=1.1B2026.06 | 237 | — | — | — | — | 5.02 | |
| ADM-GParadigm=Diffusion, Space=Pixel-Space, Params=554M2026.06 | 186.7 | — | — | — | — | 4.59 | |
| JetFormerParadigm=Autoregressive, Space=Pixel-Space, Params=2.8B2026.06 | — | — | — | — | — | 6.64 | |
| SiD2Paradigm=Diffusion, Space=Pixel-Space, Params=N/A2026.06 | — | — | — | — | — | 1.38 |