List your company

Class-conditional Image Generation on ImageNet 256x256 (test)

1.07FIDLightningDiT+IG
1.00121.46561.932.3944Oct 26, 2023Mar 24, 2024Aug 21, 2024Jan 18, 2025Jun 17, 2025Nov 14, 2025Apr 14, 2026

Evaluation results

MethodLinks
LightningDiT+IGEpochs=680, #Params=678M, CFG=true, Sampling Protocol=Balanced sampling2025.121.07274.17966——
RAE (DiTDH)Epochs=800, #Params=839M, CFG=true, Sampling Protocol=Balanced sampling2025.121.13262.67867——
RAE (DiTDH-XL/2)Type=Flow, Tokenizer=C, Loss=MSE, #Param=839M2026.041.13262.6————
REPA-EEpochs=800, #Params=675M, CFG=true, Sampling Protocol=Balanced sampling2025.121.153047966——
xAREpochs=800, #Params=1.1B, CFG=true, Sampling Protocol=Balanced sampling2025.121.24301.68364——
LightningDiT+IGEpochs=680, #Params=678M, CFG=false, Sampling Protocol=Balanced sampling2025.121.24229.67866——
BitDance-H-1xType=Hybrid, Tokenizer=D, Loss=MSE, #Param=1B2026.041.24304.4————
DE-CMNFE=50, #Params=675M, Time=0.582026.021.26—————
DDTEpochs=400, #Params=675M, CFG=true, Sampling Protocol=Balanced sampling2025.121.26310.67965——
LightningDiTEpochs=800, #Params=675M, CFG=true, Sampling Protocol=Balanced sampling2025.121.28300.58065——
REPAEpochs=800, #Params=675M, CFG=true, Sampling Protocol=Balanced sampling2025.121.29306.37964——
OURS, FLOP HEAVYPARAMS=399M, NFE=2562026.031.32294.274643.71—
DE-CMNFE=2, #Params=675M, Time=0.0242026.021.33—————
LightningDiT-XL/1NFE=250 × 2, #Params=675M, Time=> 2.82, CFG weight=6.72026.021.35—————
SiD2, FLOP HEAVYPARAMS=397M, NFE=5122026.031.37286.373653.83—
REPA-XL/2NFE=250 × 2, #Params=675M, Time=> 2.89, CFG weight=1.82026.021.42—————
OURS, SMALLPARAMS=399M, NFE=2562026.031.4229773653.82—
REPA (SiT-XL/2)Type=Flow, Tokenizer=C, Loss=MSE, #Param=675M2026.041.42305.7————
SiT-XL/2 + REPA, CFG=1.5Train Steps=4M, Params (M)=675, GFlops/Iter=140.5, Stage 2 PFlops=143.9K, Total PFlops=143.9K2025.101.42—————
RAE (DiTDH)Epochs=800, #Params=839M, CFG=false, Sampling Protocol=Balanced sampling2025.121.51242.97963——
MARNFE=100, #Params=945M2026.021.55—————
MAREpochs=800, #Params=943M, CFG=true, Sampling Protocol=Balanced sampling2025.121.55303.78162——
MAR-HLatency/s=9.97, Speedup=1.002025.121.55303.7————
TransDiff-HLatency/s=6.72, Speedup=1.002025.121.55297.9————
MARType=Hybrid, Tokenizer=C, Loss=MSE, #Param=943M2026.041.55303.7————
FACMNFE=4, #Params=675M, Time=0.0472026.021.56—————
TransDiff-H + Fast-ARDiffLatency/s=1.38, Speedup=4.882025.121.6295————
MAR-H + Fast-ARDiffLatency/s=2.23, Speedup=4.472025.121.61304.3————
TransDiff-LLatency/s=3.17, Speedup=1.002025.121.61295.1————
PIXELDIT-XLPARAMS=797M, NFE=1002026.031.61292.778644.68—
TransDiff-L + Fast-ARDiffLatency/s=0.68, Speedup=4.692025.121.66294.1————
TransDiff-H + SDLatency/s=3.52, Speedup=1.912025.121.68291.1————
SiD2, SMALLPARAMS=397M, NFE=5122026.031.68288.272654.04—
MAR-H + LazyMARLatency/s=4.24, Speedup=2.352025.121.69299.2————
DiT-XL/2 with CADSGuidance scale (w_CFG)=22023.101.7268.777764——
DE-CMNFE=1, #Params=675M, Time=0.0112026.021.7—————
REPA-EEpochs=800, #Params=675M, CFG=false, Sampling Protocol=Balanced sampling2025.121.7217.37766——
TransDiff-H + DMDLatency/s=3.48, Speedup=1.932025.121.71289.9————
VAREpochs=350, #Params=2.0B, CFG=true, Sampling Protocol=Balanced sampling2025.121.73350.28260——
MAR-H + DMDLatency/s=4.17, Speedup=2.392025.121.73301————
MeanFlow-†NFE=2, #Params=675M, Time=0.048, distillation_reproduction=true2026.021.74—————
FACMNFE=1, #Params=675M, Time=0.0112026.021.76—————
MAR-L + Fast-ARDiffLatency/s=1.21, Speedup=4.382025.121.77297.1————
MAR-LLatency/s=5.31, Speedup=1.002025.121.78296————
MDT-G2023.101.79283.018161——
TransDiff-L + DMDLatency/s=1.61, Speedup=1.972025.121.79288.3————
DiT-XL/2 with CADSGuidance scale (w_CFG)=2.252023.101.81288.17864——
MAR-L + DMDLatency/s=1.99, Speedup=2.672025.121.81295.5————
GRN-GType=AR, Tokenizer=D, Loss=CE, #Param=2B2026.041.81299————
JIT-GPARAMS=2B2026.031.82292.67962——
JiT-G/16Type=Flow, Tokenizer=C, Loss=MSE, #Param=2B2026.041.82292.6————
DiT-G++2023.101.83281.537864——
JiT-H/16Type=Flow, Tokenizer=C, Loss=MSE, #Param=953M2026.041.86303.4————
TransDiff-L + SDLatency/s=1.75, Speedup=1.812025.121.88283.6————
RepTok-L + CFG=1.5Stage 1 PFlops=30.4K, Train Steps=460K, Params (M)=516, GFlops/Iter=25.0, Stage 2 PFlops=11.7K, Total PFlops=42.1K2025.101.88—————
VAR-d30NFE=10, #Params=2B, Time=12026.021.92—————
VAREpochs=350, #Params=2.0B, CFG=false, Sampling Protocol=Balanced sampling2025.121.92323.18259——
VAR-d30Type=AR, Tokenizer=D, Loss=CE, #Param=2B2026.041.92323.1————
DiT-XL/2 with CADSGuidance scale (w_CFG)=2.52023.101.93297.967863——
MAR-L + LazyMARLatency/s=2.29, Speedup=2.322025.121.93297.4————
sCMs-†NFE=2, #Params=675M, Time=0.024, distillation_reproduction=true2026.021.94—————
PIXELFLOW-XLPARAMS=677M, NFE=ADAPT.2026.031.98282.181605.83—
SiT-XL/2Type=Flow, Tokenizer=C, Loss=MSE, #Param=675M2026.042.06277.5————
GRN-HType=AR, Tokenizer=D, Loss=CE, #Param=952M2026.042.06316.1————
SiT-XL/2 + CFG=1.5Train Steps=7M, Params (M)=675, GFlops/Iter=118.6, Stage 2 PFlops=212.5K, Total PFlops=212.5K2025.102.06—————
RepTok-LStage 1 PFlops=30.4K, Train Steps=460K, Params (M)=516, GFlops/Iter=25.0, Stage 2 PFlops=11.7K, Total PFlops=42.1K2025.102.06—————
VAR-d24Type=AR, Tokenizer=D, Loss=CE, #Param=1B2026.042.09312.9————
L-DiTNFE=250, #Params=3B, Time=4.12026.022.1—————
LightningDiTEpochs=800, #Params=675M, CFG=false, Sampling Protocol=Balanced sampling2025.122.112077765——
VDM++2023.102.12267.7————
SiT-XL/2NFE=250 × 2, #Params=675M, Time=> 2.89, CFG weight=1.52026.022.15—————
PIXNERD-XL/16PARAMS=700M, NFE=1002026.032.1529779594.55—
RandAR-XXLType=AR, Tokenizer=D, Loss=CE, #Param=1.4B2026.042.15322————
LightningDiT-XL/1NFE=250, #Params=675M, Time=2.822026.022.17—————
LlamaGen-3BType=AR, Parameters=3.1B, cfg=1.652024.062.18263.338158——
MeanFlowNFE=2, #Params=675M, Time=0.0242026.022.2—————
VA-VAEEpochs=800, Steps=2502026.022.23————74.3
DiT-XL/2-GGuidance scale (w_CFG)=1.52023.102.27278.248357——
DiT-XL/2Type=Diffusion, Parameters=675M2024.062.27278.28357——
DiT-XL/2NFE=250 × 2, #Params=675M, Time=> 3.75, CFG weight=1.52026.022.27—————
DiT-XL/2Type=Diff, Params=675M, Steps=2502025.122.27278.28357——
DiT-XL/2Type=Diffusion, Tokenizer=C, Loss=MSE, #Param=675M2026.042.27278.2————
StyleGAN-XL2023.102.3265.127853——
StyleGan-XLType=GAN, Parameters=166M2024.062.3265.17853——
StyleGAN-XLNFE=1, #Params=166M, Time=0.32026.022.3—————
StyleGAN-XLType=GAN, Params=166M, Steps=12025.122.3265.17853——
LlamaGen-3BType=AR, Parameters=3.1B, cfg=1.752024.062.32280.18256——
LlamaGen-XXLType=AR, Parameters=1.4B, cfg=1.752024.062.34253.98059——
LlamaGen-XXLType=AR, Tokenizer=D, Loss=CE, #Param=1.4B2026.042.34253.9————
MAREpochs=800, #Params=943M, CFG=false, Sampling Protocol=Balanced sampling2025.122.35227.87962——
JiT-L/16Type=Flow, Tokenizer=C, Loss=MSE, #Param=459M2026.042.36298.5————
LV-RAEEpochs=800, Steps=2502026.022.42————58.2
SimpleDiffusion2023.102.44256.3————
VAR-d20Type=AR, Tokenizer=D, Loss=CE, #Param=600M2026.042.57302.6————
LlamaGen-XLType=AR, Parameters=775M, cfg=1.752024.062.62244.088057——
LlamaGen-XLType=AR, Tokenizer=D, Loss=CE, #Param=775M2026.042.62244.1————
GRN-LType=AR, Tokenizer=D, Loss=CE, #Param=458M2026.042.64314.8————
NAR-XLType=Parallelized AR, Params=816M, Steps=312025.122.7277.58158——
TFM-XL/2# Params.=676M, NFE=22026.032.77—————
MeanFlow-†NFE=1, #Params=675M, Time=0.041, distillation_reproduction=true2026.022.79—————