Image Reconstruction on ImageNet (val)
0.011rFIDFlux VAE
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Flux VAEScale Factor=4.02026.03 | 0.011 | 0.991 | 41.71 | 0.0034 | — | — | 0.68 | — | — | — | |
| Flux VAEScale Factor=2.02026.03 | 0.012 | 0.99 | 41.51 | 0.0061 | — | — | 0.71 | — | — | — | |
| Flux VAEScale Factor=3.02026.03 | 0.012 | 0.99 | 41.6 | 0.0061 | — | — | 0.7 | — | — | — | |
| Flux VAEScale Factor=1.52026.03 | 0.033 | 0.977 | 37.18 | 0.0081 | — | — | 1.96 | — | — | — | |
| FLUX.1[dev]-VAETokenizer Category=Specialized, Resolution=512, Note=re-evaluated using official checkpoints2026.03 | 0.06 | 0.93 | 33.65 | — | — | — | — | — | — | — | |
| FLUX.1[dev]-VAERes.=5122026.04 | 0.06 | 0.93 | 33.65 | — | — | — | — | — | — | — | |
| Flux VAEScale Factor=0.52026.03 | 0.069 | 0.927 | 30.82 | 0.0126 | — | — | 8.31 | — | — | — | |
| Flux VAEScale Factor=1.02026.03 | 0.112 | 0.914 | 28.28 | 0.0184 | — | — | 15.05 | — | — | — | |
| Distilled VAEScale Factor=4.02026.03 | 0.115 | 0.983 | 36.2 | 0.0041 | — | — | 2.4 | — | — | — | |
| Tuna-RRes.=5122026.04 | 0.12 | 0.93 | 32.22 | — | — | — | — | — | — | — | |
| Tuna-2Res.=5122026.04 | 0.15 | 0.93 | 32.8 | — | — | — | — | — | — | — | |
| FLUX-VAETokenizer Category=Generation-oriented Tokenizer2026.01 | 0.176 | 0.917 | 32.86 | 0.044 | — | — | — | — | — | — | |
| PS-VAERes.=2562026.04 | 0.2 | 0.82 | 28.79 | — | — | — | — | — | — | — | |
| SD3-VAETokenizer Category=Generation-oriented Tokenizer2026.01 | 0.201 | 0.886 | 31.29 | 0.059 | — | — | — | — | — | — | |
| SD3-VAESemantics Distillation=false2025.07 | 0.21 | — | — | — | — | — | — | — | — | — | |
| OpenVision 3Tokenizer Category=Unified Tokenizer2026.01 | 0.216 | 0.885 | 30.33 | 0.061 | — | — | — | — | — | — | |
| DC-AETokenizer Category=Specialized, Resolution=5122026.03 | 0.22 | 0.71 | 26.15 | — | — | — | — | — | — | — | |
| DC-AERes.=5122026.04 | 0.22 | 0.71 | 26.15 | — | — | — | — | — | — | — | |
| VA-VAETokenizer Category=Specialized, Resolution=2562026.03 | 0.26 | 0.8 | 28.59 | — | — | — | — | — | — | — | |
| VA-VAERes.=2562026.04 | 0.26 | 0.8 | 28.59 | — | — | — | — | — | — | — | |
| AlignTokf=14, #Tok.=256, Dim.=322026.05 | 0.26 | — | 25.8 | — | — | — | — | — | — | — | |
| Distilled VAEScale Factor=3.02026.03 | 0.268 | 0.985 | 35.78 | 0.0046 | — | — | 2.64 | — | — | — | |
| VA-VAESemantics Distillation=true2025.07 | 0.28 | — | — | — | — | — | — | — | — | — | |
| VA-VAEf=16, #Tok.=256, Dim.=322026.05 | 0.28 | 0.79 | 28 | — | — | — | — | — | — | — | |
| VMFAE(DINOv2)f=16, #Tok.=256, Dim.=322026.05 | 0.29 | 0.82 | 25.5 | — | — | — | — | 225.6 | — | — | |
| VFMAE(SigLIP2)f=16, #Tok.=256, Dim.=322026.05 | 0.29 | 0.79 | 24.6 | — | — | — | — | 225.4 | — | — | |
| UniTokTokenizer Category=Unified Tokenizer2026.01 | 0.362 | 0.742 | 25.34 | 0.132 | — | — | — | — | — | — | |
| UniTokTokenizer Category=Unified, Resolution=2562026.03 | 0.38 | — | — | — | — | — | — | — | — | — | |
| UniComTokenizer Category=Unified, Resolution=1024, d (feature dimension)=11522026.03 | 0.38 | 0.61 | 22.6 | — | — | — | — | — | — | — | |
| UniTokRes.=2562026.04 | 0.38 | — | — | — | — | — | — | — | — | — | |
| UniComTokenizer Category=Unified, Resolution=1024, d (feature dimension)=642026.03 | 0.42 | 0.61 | 22.28 | — | — | — | — | — | — | — | |
| GQbpp (K x N)=0.28 (2^18 x 1024), ImageNet Trained Epochs=40, Params(M)=872025.12 | 0.424 | 0.804 | 27.86 | 0.054 | — | — | — | — | — | — | |
| Emu3.5-IBQ#Tokens=1,024, Codebook=131,072, BPP=0.0664, Resolution=512 x 5122026.05 | 0.46 | — | 23.82 | — | — | — | — | — | — | — | |
| MAETokSemantics Distillation=true2025.07 | 0.48 | — | — | — | — | — | — | — | — | — | |
| GQbpp (K x N)=0.22 (2^14 x 1024), ImageNet Trained Epochs=40, Params(M)=872025.12 | 0.491 | 0.762 | 25.31 | 0.064 | — | — | — | — | — | — | |
| GigaTokTokenizer Category=Specialized, Resolution=2562026.03 | 0.51 | 0.69 | 21.32 | — | — | — | — | — | — | — | |
| GigaTokRes.=2562026.04 | 0.51 | 0.69 | 21.32 | — | — | — | — | — | — | — | |
| VFMVAEf=14, #Tok.=256, Dim.=322026.05 | 0.52 | 0.59 | 23 | — | — | — | — | 214.1 | — | — | |
| MAR-VAESemantics Distillation=false2025.07 | 0.53 | — | — | — | — | — | — | — | — | — | |
| MingTokTokenizer Category=Unified, Resolution=512, Note=re-evaluated using official checkpoints2026.03 | 0.53 | 0.61 | 23.49 | — | — | — | — | — | — | — | |
| MingTokRes.=5122026.04 | 0.53 | 0.61 | 23.49 | — | — | — | — | — | — | — | |
| FlowMo-Hibpp (K x N)=0.22 (2^14 x 1024), ImageNet Trained Epochs=300, Params(M)=9452025.12 | 0.56 | 0.785 | 24.93 | 0.073 | — | — | — | — | — | — | |
| Distilled VAEScale Factor=2.02026.03 | 0.563 | 0.954 | 33.93 | 0.0334 | — | — | 4.08 | — | — | — | |
| RAEf=14, #Tok.=256, Dim.=7682026.05 | 0.57 | 0.61 | 18.8 | — | — | — | — | 226.8 | — | — | |
| VAR#Tokens=2,240, Codebook=4,096, BPP=0.1025, Resolution=512 x 5122026.05 | 0.58 | — | 23.43 | — | — | — | — | — | — | — | |
| SD-VAEf=16, #Tok.=256, Dim.=162026.05 | 0.58 | 0.72 | 25.7 | — | — | — | — | — | — | — | |
| Llama-Gen-32bpp (K x N)=0.22 (2^14 x 1024), ImageNet Trained Epochs=40, Params(M)=702025.12 | 0.59 | 0.768 | 24.44 | 0.064 | — | — | — | — | — | — | |
| SD-VAETokenizer Category=Generation-oriented Tokenizer2026.01 | 0.606 | 0.745 | 26.26 | 0.133 | — | — | — | — | — | — | |
| SD-VAESemantics Distillation=false2025.07 | 0.61 | — | — | — | — | — | — | — | — | — | |
| RAERes.=2562026.04 | 0.61 | 0.44 | 19.2 | — | — | — | — | — | — | — | |
| Idealf=16, Size=16384, Dim.=64, #Res.=3842026.06 | 0.61 | — | — | — | 100 | — | — | 230.4 | — | — | |
| MAE-TokTokenizer Category=Specialized, Resolution=5122026.03 | 0.62 | — | — | — | — | — | — | — | — | — | |
| MAE-TokRes.=5122026.04 | 0.62 | — | — | — | — | — | — | — | — | — | |
| TokenFlowTokenizer Category=Unified, Resolution=3842026.03 | 0.63 | 0.73 | 22.77 | — | — | — | — | — | — | — | |
| TokenFlowRes.=3842026.04 | 0.63 | 0.73 | 22.77 | — | — | — | — | — | — | — | |
| SD-VAEType=cont., #tok=10242026.03 | 0.64 | — | — | — | — | 223.8 | — | — | — | — | |
| IBQ-16k#Tokens=1,024, Codebook=16,384, BPP=0.0547, Resolution=512 x 5122026.05 | 0.64 | — | 23.39 | — | — | — | — | — | — | — | |
| O-MAGVIT2-262k#Tokens=1,024, Codebook=262,144, BPP=0.0703, Resolution=512 x 5122026.05 | 0.65 | — | 23.98 | — | — | — | — | — | — | — | |
| FAEf=14, #Tok.=256, Dim.=642026.05 | 0.66 | — | — | — | — | — | — | — | — | — | |
| l-DeTokSemantics Distillation=false2025.07 | 0.68 | — | — | — | — | — | — | — | — | — | |
| LlamaGen-T2I#Tokens=1,024, Codebook=16,384, BPP=0.0547, Resolution=512 x 5122026.05 | 0.69 | — | 22.8 | — | — | — | — | — | — | — | |
| InsightTok#Tokens=1,024, Codebook=16,384, BPP=0.0547, Resolution=512 x 5122026.05 | 0.69 | — | 23.64 | — | — | — | — | — | — | — | |
| ImageFolderf=-, Size=32768, Dim.=32, #Tok.=2862026.05 | 0.69 | — | — | — | — | — | — | 201.5 | 100 | — | |
| ImageFolderf=-, Size=32768, Dim.=32, #Res.=2562026.06 | 0.69 | — | — | — | 100 | — | — | 201.5 | — | — | |
| TexTokTokenizer Category=Specialized, Resolution=5122026.03 | 0.73 | 0.66 | 24.45 | — | — | — | — | — | — | — | |
| TexTokRes.=5122026.04 | 0.73 | 0.66 | 24.45 | — | — | — | — | — | — | — | |
| NVG#Tokens=511*, Ratio=16, Codebook=4,0962025.08 | 0.74 | — | — | 0.1875 | 100 | — | — | — | — | — | |
| Distilled VAEScale Factor=1.52026.03 | 0.822 | 0.929 | 31.33 | 0.0395 | — | — | 7.43 | — | — | — | |
| l-DeTok + DistillationSemantics Distillation=true2025.07 | 0.85 | — | — | — | — | — | — | — | — | — | |
| VFMTokf=-, Size=16384, Dim.=12, #Tok.=2562026.05 | 0.89 | — | — | — | — | — | — | 215.4 | 100 | — | |
| VFMTokf=-, Size=16384, Dim.=12, #Res.=3362026.06 | 0.89 | — | — | — | 100 | — | — | 215.4 | — | — | |
| OptVQLatent Size=16 x 16 x 8, From Scratch=true, #Tokens=16,3842024.12 | 0.91 | 0.729 | 27.57 | 0.066 | — | — | — | — | — | — | |
| VARf=16, Size=4096, Dim.=32, #Tok.=6802026.05 | 0.92 | — | — | — | — | — | — | 196 | — | 100 | |
| VARf=16, Size=4096, Dim.=32, #Res.=2562026.06 | 0.92 | — | — | — | 100 | — | — | 196 | — | — | |
| Chameleon#Tokens=1,024, Codebook=8,192, BPP=0.0508, Resolution=512 x 5122026.05 | 0.93 | — | 22.53 | — | — | — | — | — | — | — | |
| Wan2.1-VAETokenizer Category=Generation-oriented Tokenizer2026.01 | 0.945 | 0.886 | 31.34 | 0.058 | — | — | — | — | — | — | |
| VQGANf=8, #Tok.=576, Ref.=[72]2026.05 | 0.95 | — | — | — | — | — | — | 197.3 | — | 99.7 | |
| LlamaGenf=16, Size=16384, Dim.=8, #Res.=3842026.06 | 0.95 | — | — | — | 99.7 | — | — | 197.3 | — | — | |
| CosmosTokenizer Category=Generation-oriented Tokenizer2026.01 | 0.959 | 0.7 | 25.07 | 0.167 | — | — | — | — | — | — | |
| BSQbpp (K x N)=0.28 (2^18 x 1024), ImageNet Trained Epochs=200, Params(M)=1752025.12 | 0.99 | 0.817 | 27.78 | 0.063 | — | — | — | — | — | — | |
| OptVQLatent Size=16 x 16 x 4, From Scratch=true, #Tokens=16,3842024.12 | 1 | 0.717 | 26.59 | 0.076 | — | — | — | — | — | — | |
| IBQ#Tokens=16 × 16, Ratio=16, Codebook=262,1442025.08 | 1 | — | — | 0.203 | 84 | — | — | — | — | — | |
| LooC-VAEK x d*=1024 x 42026.01 | 1.01 | 0.716 | 29.15 | — | 100 | — | — | — | — | — | |
| LooC-VAECodebook size (K x d*)=1024 × 42026.01 | 1.01 | 0.716 | 29.15 | 0.716 | 100 | — | — | — | — | — | |
| Distilled VAEScale Factor=0.52026.03 | 1.031 | 0.859 | 26.99 | 0.0478 | — | — | 20.04 | — | — | — | |
| GQbpp (K x N)=0.07 (2^14 x 256), ImageNet Trained Epochs=40, Params(M)=872025.12 | 1.04 | 0.642 | 22.3 | 0.116 | — | — | — | — | — | — | |
| TiTokf=-, Size=8192, Dim.=64, #Tok.=2562026.05 | 1.05 | — | — | — | — | — | — | 191.5 | 100 | — | |
| TiTokf=-, Size=8192, Dim.=64, #Res.=2562026.06 | 1.05 | — | — | — | 100 | — | — | 191.5 | — | — | |
| RAE (CLIP)Tokenizer Category=Unified Tokenizer2026.01 | 1.06 | 0.403 | 17.44 | 0.324 | — | — | — | — | — | — | |
| VAR#Tokens=680, Ratio=16, Codebook=4,0962025.08 | 1.06 | — | — | 0.1863 | 100 | — | — | — | — | — | |
| SD-VAETokenizer Category=Specialized, Resolution=2562026.03 | 1.06 | 0.86 | 28.62 | — | — | — | — | — | — | — | |
| SD-VAERes.=2562026.04 | 1.06 | 0.86 | 28.62 | — | — | — | — | — | — | — | |
| TokenBridgebpp (K x N)=0.375 (2^6 x 4096), Params(M)=832025.12 | 1.11 | — | — | — | — | — | — | — | — | — | |
| Mo-VQGANLatent Size=16 x 16 x 4, From Scratch=true, #Tokens=1,0242024.12 | 1.12 | 0.673 | 22.42 | 0.113 | — | — | — | — | — | — | |
| MoVQK x d*=1024 x 642026.01 | 1.12 | 0.6731 | 22.42 | — | 63 | — | — | — | — | — | |
| MoVQCodebook size (K x d*)=1024 × 642026.01 | 1.12 | 0.6731 | 22.42 | 0.1132 | 63 | — | — | — | — | — | |
| VQGAN-SDbpp (K x N)=0.22 (2^14 x 1024), ImageNet Trained Epochs=OpenImages, Params(M)=832025.12 | 1.14 | — | — | — | — | — | — | — | — | — | |
| DINO-Tokf=16, Size=16384×2, Dim.=832, #Res.=2562026.06 | 1.15 | — | — | — | — | — | — | — | — | — | |
| OpenMagViT-V2bpp (K x N)=0.07 (2^14 x 256), ImageNet Trained Epochs=300, Params(M)=1702025.12 | 1.17 | 0.64 | 21.63 | 0.111 | — | — | — | — | — | — | |
| Open-MAGVIT2#Tokens=16 × 16, Ratio=16, Codebook=262,1442025.08 | 1.17 | — | — | 0.2038 | 100 | — | — | — | — | — |