List your company

Class-Conditional Image Generation on ImageNet 256x256 (test val)

1.55FIDMAR-H
0.0889.956519.82529.6935Feb 18, 2021Nov 28, 2021Sep 7, 2022Jun 17, 2023Mar 26, 2024Jan 3, 2025Oct 14, 2025

Evaluation results

MethodLinks
MAR-HType=MAR, #Para.=943M2024.121.55—303.70.810.62
MAR-LType=MAR, #Para.=479M2024.121.78—2960.810.6
MDT-XL/2-Gdynamic CFG=true2023.091.794.57283.018161
DoD-XL-GStage=S2, cfg=3.5, Images=256M, Params=613M2024.101.835263.690.770.65
RDMCFG=3.5, class balance=true2023.091.873.97278.758159
VAR-d30Type=VAR, #Para.=2.0B2024.121.92—323.10.820.59
Flag-DiT-GImages=256M, Params=4.23B2024.101.964.43284.80.820.61
RDMCFG=3.52023.091.993.99260.458158
Efficient-DiT-Gcfg=1.5, Images=675M, Params=3.01B2024.102.014.49271.040.820.6
IBQ-XXLType=Vanilla AR, #Para.=2.1B2024.122.05—286.730.830.57
VAR-d24Type=VAR, #Para.=1.0B2024.122.09—312.90.820.59
IBQ-XLType=Vanilla AR, #Para.=1.1B2024.122.14—278.990.830.56
SiT-XL-Gcfg=1.5, Images=1792M, Params=675M2024.102.154.5258.090.810.6
FiTv2-3B-Gcfg=1.5, Images=256M, Params=3B2024.102.154.49276.320.820.59
LlamaGen-3B (upsampled)Type=Vanilla AR, #Para.=3.1B, Note=Generated at 384x384, resized to 256x2562024.122.18—263.330.810.58
MDT-XL/2-GCFG=1.3252023.092.264.28246.068159
FiTv2-XL-Gcfg=1.5, Images=512M, Params=671M2024.102.264.53260.950.810.59
DiT-XL/2-GCFG=1.52023.092.274.6278.248357
DiT-XL-Gcfg=1.5, Images=1792M, Params=675M2024.102.274.6278.240.830.57
DiT-XL/2Type=Diffusion, #Para.=675M2024.122.27—278.20.830.57
StyleGAN-XL2023.092.34.02265.127853
StyleGAN-XL2024.102.34.02265.120.780.53
MAR-BType=MAR, #Para.=208M2024.122.31—281.70.820.57
Open-MAGVIT2-XLType=Vanilla AR, #Para.=1.5B2024.122.33—271.770.840.54
DoD-XL-GStage=S2, cfg=3.5, Images=102.4M, Params=613M2024.102.345228.420.780.61
LlamaGen-XXL (upsampled)Type=Vanilla AR, #Para.=1.4B, Note=Generated at 384x384, resized to 256x2562024.122.34—253.90.80.59
U-ViT-H-Gcfg=1.4, Images=512M, Params=501M2024.102.355.68265.020.820.57
IBQ-LType=Vanilla AR, #Para.=649M2024.122.45—267.480.830.52
Open-MAGVIT2-LType=Vanilla AR, #Para.=804M2024.122.51—271.70.840.54
VAR-d20Type=VAR, #Para.=600M2024.122.57—302.60.830.56
LlamaGen-XL (upsampled)Type=Vanilla AR, #Para.=775M, Note=Generated at 384x384, resized to 256x2562024.122.62—244.080.80.57
IBQ-BType=Vanilla AR, #Para.=342M2024.122.88—254.730.840.51
ViT-VQGAN-reType=Vanilla AR, #Para.=1.7B2024.123.04—227.4——
LlamaGen-3BType=Vanilla AR, #Para.=3.1B2024.123.06—279.720.840.53
LlamaGen-L (upsampled)Type=Vanilla AR, #Para.=343M, Note=Generated at 384x384, resized to 256x2562024.123.07—256.060.830.52
Open-MAGVIT2-BType=Vanilla AR, #Para.=343M2024.123.08—258.260.850.51
LlamaGen-XXLType=Vanilla AR, #Para.=1.4B2024.123.09—253.610.830.53
DoD-B-GStage=S2, cfg=5.5, Images=384M, Params=191M2024.103.155.74248.480.820.54
VAR-d16Type=VAR, #Para.=310M2024.123.3—274.40.840.51
LlamaGen-XLType=Vanilla AR, #Para.=775M2024.123.39—227.080.810.54
DoD-B-GStage=S2, cfg=5.5, Images=256M, Params=191M2024.103.425.65236.590.820.52
LDM-4-GCFG=1.52023.093.6—247.678748
LDM-4-Gcfg=1.5, Images=214M, Params=395M2024.103.65.12247.670.870.48
LDM-4Type=Diffusion, #Para.=400M2024.123.6—247.7——
RQTran.-reType=Vanilla AR, #Para.=3.8B2024.123.8—323.7——
LlamaGen-LType=Vanilla AR, #Para.=343M2024.123.8—248.280.830.51
ADM-U,G2023.093.946.14215.848353
ADM-G,UImages=507M, Params=673M2024.103.946.14215.840.830.53
ViT-VQGANType=Vanilla AR, #Para.=1.7B2024.124.17—175.1——
FiT-XL-Gcfg=1.5, Images=512M, Params=824M2024.104.2110.01254.870.840.51
CDM2023.094.88—158.71——
CDM2024.104.88—158.71——
CDMType=Diffusion2024.124.88—158.7——
DoD-B-GStage=S2, cfg=5.5, Images=102.4M, Params=191M2024.104.935.61187.140.810.48
VQGAN-reType=Vanilla AR, #Para.=1.4B2024.125.2—280.3——
RDM2023.095.274.39153.437562
MaskGITImages=355M2024.106.18—182.10.80.51
MDT-XL/22023.096.235.23143.027165
BigGAN-deep2023.096.957.36171.18728
BigGAN-deep2024.106.957.36171.40.870.28
BigGAN-deepTruncation=best FID2021.027————
RQTran.Type=Vanilla AR, #Para.=3.8B2024.127.55—134——
BigGANTruncation=best FID2021.027.7————
Shortcut-XL/2NFE=4, Epochs=250, #Params=84M + 675M2025.107.8————
IMMNFE=1, Epochs=6395, #Params=84M + 675M2025.108.05————
EPG-L/16NFE=1, Epochs=560, #Params=540M2025.108.82————
DiT-XL/22023.099.626.85121.56767
DoD-S-GStage=S2, cfg=5.5, Images=384M, Params=48M2024.109.716.16123.630.750.49
LDM-42023.0910.56—103.497162
Shortcut-XL/2NFE=1, Epochs=250, #Params=84M + 675M2025.1010.6————
ADM2023.0910.946.02100.986963
ADMType=Diffusion, #Para.=554M2024.1210.94—1010.690.63
DoD-S-GStage=S2, cfg=5.5, Images=256M, Params=48M2024.1011.976.23108.950.740.48
Improved DiffusionProcessing stages=two-stage2021.0212.3————
VQGANType=Vanilla AR, #Para.=1.4B2024.1215.78—74.3——
VQGANType=Vanilla AR, #Para.=227M2024.1218.65—80.40.780.26
DoD-S-GStage=S2, cfg=5.5, Images=102.4M, Params=48M2024.1019.78.0874.940.670.48
iCT-XL/2NFE=2, #Params=84M + 675M2025.1020.3————
Improved DiffusionProcessing stages=single-stage2021.0231.5————
iCT-XL/2NFE=1, #Params=84M + 675M2025.1034.24————
VQ-VAE-2Processing stages=two-stage2021.0238.1————