List your company

Text-to-Image Generation on FLUX.1 (dev)

1.0527Image RewardFresco
0.6258840.7366920.84750.958308Jan 12, 2026

Evaluation results

MethodLinks
FrescoNFE=30, Latency(s)=9.2, Speed (Latency)=2.81×, FLOPs(T)=1295.99, Speed (FLOPs)=2.87×2026.011.052732.521————
FrescoNFE=18, Latency(s)=5.72, Speed (Latency)=4.51×, FLOPs(T)=788.03, Speed (FLOPs)=4.72×2026.011.036932.581————
SVD-CacheEfficient Attention=true, N=42026.011.014732.848.093.19967.913.84
SVD-CacheEfficient Attention=true, N=52026.011.012332.9837.623.38893.544.16
FrescoNFE=11, Latency(s)=3.43, Speed (Latency)=7.52×, FLOPs(T)=486.85, Speed (FLOPs)=7.64×2026.011.00932.37————
SVD-CacheEfficient Attention=true, N=62026.011.005733.0276.813.79744.815
TaylorSeerEfficient Attention=true, N=3, O=22026.010.998932.4139.892.611,320.072.82
SpecaEfficient Attention=true, Nmax=8, Nmin=22026.010.998532.2777.423.48791.384.7
DuCaEfficient Attention=true, N=52026.010.995532.2418.183.15978.763.8
SVD-CacheEfficient Attention=true, N=72026.010.993833.1446.434.01670.445.55
ChipmunkEfficient Attention=true2026.010.993632.54812.722.021,505.872.47
FLUX.1 [dev]Efficient Attention=true, Steps=50 steps2026.010.989832.40425.8213,719.51
FoCaEfficient Attention=true, N=32026.010.98932.5779.282.781,327.212.8
TaylorSeerNFE=50, N'=3, Latency(s)=9.89, Speed (Latency)=2.61×, FLOPs(T)=1320.07, Speed (FLOPs)=2.82×2026.010.988932.413————
TaylorSeerNFE=50, N'=4, Latency(s)=9.24, Speed (Latency)=2.80×, FLOPs(T)=967.91, Speed (FLOPs)=3.84×2026.010.985732.413————
TaylorSeerEfficient Attention=true, N=4, O=22026.010.985732.4139.242.8967.913.84
DuCaNFE=50, N'=5, Latency(s)=8.18, Speed (Latency)=3.15×, FLOPs(T)=978.76, Speed (FLOPs)=3.80×2026.010.985532.241————
CluscaEfficient Attention=true, N=4, O=2, K=162026.010.98532.4419.252.791,045.583.56
FrescoNFE=9, Latency(s)=2.54, Speed (Latency)=10.16×, FLOPs(T)=362.17, Speed (FLOPs)=10.27×2026.010.982532.426————
SpargeAttentionNFE=50, Latency(s)=15.46, Speed (Latency)=1.67×, FLOPs(T)=2150.02, Speed (FLOPs)=1.73×2026.010.981232.391————
ToCaEfficient Attention=false, N=62026.010.980232.08313.161.96924.34.02
FORAEfficient Attention=true, N=32026.010.977632.26610.162.541,320.072.82
SVD-CacheEfficient Attention=true, N=82026.010.976932.8484.995.17596.076.24
TaylorSeerEfficient Attention=true, N=5, O=22026.010.976832.4677.463.46893.544.16
FoCaEfficient Attention=true, N=42026.010.975732.5389.352.761,050.73.54
DuCaEfficient Attention=true, N=72026.010.975732.0666.743.83760.144.89
Bottleneck SamplingNFE=30, Latency(s)=11.32, Speed (Latency)=2.28×, FLOPs(T)=1582.77, Speed (FLOPs)=2.35×2026.010.973932.221————
FLUX.1-devNFE=50, Latency(s)=25.82, Speed (Latency)=1.00×, FLOPs(T)=3719.5, Speed (FLOPs)=1.00×2026.010.973632.404————
FORAEfficient Attention=true, N=42026.010.97332.1428.123.14967.913.84
CluscaEfficient Attention=true, N=5, O=1, K=162026.010.971832.3197.053.66897.034.14
FoCaEfficient Attention=true, N=62026.010.971332.9227.543.42745.394.99
CluscaEfficient Attention=true, N=6, O=1, K=162026.010.970432.2177.133.62748.484.97
ToCaNFE=50, N'=6, Latency(s)=13.16, Speed (Latency)=1.96×, FLOPs(T)=924.3, Speed (FLOPs)=4.02×2026.010.970232.083————
TaylorSeerEfficient Attention=true, N=7, O=22026.010.969832.1286.773.81671.395.54
60% stepsNFE=30, Latency(s)=16.7, Speed (Latency)=1.55×, FLOPs(T)=2231.7, Speed (FLOPs)=1.67×2026.010.966332.312————
60% stepsEfficient Attention=true2026.010.966332.31216.71.552,231.71.67
RALUNFE=30, Latency(s)=11.03, Speed (Latency)=2.34×, FLOPs(T)=1499.79, Speed (FLOPs)=2.48×2026.010.964432.198————
36% stepsNFE=18, Latency(s)=9.89, Speed (Latency)=2.61×, FLOPs(T)=1339.02, Speed (FLOPs)=2.77×2026.010.955332.114————
RALUNFE=18, Latency(s)=6.34, Speed (Latency)=4.07×, FLOPs(T)=904.98, Speed (FLOPs)=4.11×2026.010.948132.274————
34% stepsEfficient Attention=true2026.010.945332.1149.072.851,264.633.13
ToCaEfficient Attention=false, N=82026.010.945131.99311.362.27784.544.74
TeaCacheNFE=50, threshold (l)=0.25, Latency(s)=14.1, Speed (Latency)=1.83×, FLOPs(T)=1937.24, Speed (FLOPs)=1.92×2026.010.944932.247————
∆-DiTEfficient Attention=true, N=22026.010.944432.27317.81.452,480.011.5
RALUNFE=10, Latency(s)=3.73, Speed (Latency)=6.92×, FLOPs(T)=540.62, Speed (FLOPs)=6.88×2026.010.928932.113————
FLUX.1-schnellNFE=8, Latency(s)=4.21, Speed (Latency)=6.12×, FLOPs(T)=595.12, Speed (FLOPs)=6.25×2026.010.909733.837————
TeaCacheNFE=50, threshold (l)=0.8, Latency(s)=6.64, Speed (Latency)=3.89×, FLOPs(T)=892.35, Speed (FLOPs)=4.17×2026.010.880531.827————
∆-DiTEfficient Attention=true, N=32026.010.872132.10213.021.981,686.762.21
TeaCacheEfficient Attention=true, l=0.82026.010.868331.7047.213.58892.354.17
TaylorSeerNFE=50, N'=9, Latency(s)=4.85, Speed (Latency)=5.32×, FLOPs(T)=596.07, Speed (FLOPs)=6.24×2026.010.856231.653————
DuCaEfficient Attention=true, N=92026.010.838231.7597.273.55690.255.39
TeaCacheEfficient Attention=true, l=12026.010.837931.8778.193.19743.635.01
DuCaNFE=50, N'=9, Latency(s)=7.27, Speed (Latency)=3.55×, FLOPs(T)=690.25, Speed (FLOPs)=5.39×2026.010.818231.759————
FORAEfficient Attention=true, N=62026.010.77631.7428.173.16744.84.99
ToCaEfficient Attention=false, N=102026.010.715531.8087.933.25714.665.2
ToCaNFE=50, N'=10, Latency(s)=7.93, Speed (Latency)=3.25×, FLOPs(T)=714.66, Speed (FLOPs)=5.20×2026.010.705531.808————
TeaCacheNFE=50, threshold (l)=1.6, Latency(s)=3.78, Speed (Latency)=6.82×, FLOPs(T)=520.54, Speed (FLOPs)=7.15×2026.010.642331.656————