Image Reconstruction on OmniDoc-TokenBench 256x256 (test)
97.06SSIMQwen-Image-VAE-2.0-f16c128
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Qwen-Image-VAE-2.0-f16c128Setting=f16c128, Compression Factor=f16, Architecture=f16 Compression VAE2026.05 | 97.06 | 30.45 | 1.67 | 0.79 | 96.17 | |
| FLUX.2-devSetting=f16c128, Compression Factor=f16, Architecture=f16 Compression VAE2026.05 | 95.44 | 27.72 | 2.16 | 0.73 | 95.35 | |
| FLUX.1-devSetting=f8c16, Compression Factor=f8, Architecture=f8 Compression VAE2026.05 | 93.64 | 26.24 | 2.46 | 0.55 | 95.46 | |
| Qwen-Image-VAE-2.0-f16c64Setting=f16c64, Compression Factor=f16, Architecture=f16 Compression VAE2026.05 | 92.79 | 26 | 3.82 | 1.94 | 92.44 | |
| HunyuanVideoSetting=f8c16, Compression Factor=f8, Architecture=f8 Compression VAE2026.05 | 92.27 | 25.26 | 4.34 | 2.03 | 92.66 | |
| Cosmos-0.1-CI8x8Setting=f8c16, Compression Factor=f8, Architecture=f8 Compression VAE2026.05 | 90.57 | 24.29 | 4.64 | 2.89 | 90.33 | |
| Qwen-ImageSetting=f8c16, Compression Factor=f8, Architecture=f8 Compression VAE2026.05 | 89.98 | 24.94 | 5.19 | 4.48 | 90.73 | |
| Stepvideo-T2VSetting=f16c64, Compression Factor=f16, Architecture=f16 Compression VAE2026.05 | 89.7 | 23.69 | 6.5 | 6.02 | 88.38 | |
| Qwen-Image-VAE-2.0-f32c192Setting=f32c192, Compression Factor=f32, Architecture=f32 Compression VAE2026.05 | 89.08 | 23.84 | 4.97 | 1.98 | 85.55 | |
| HunyuanImage-3.0Setting=f16c32, Compression Factor=f16, Architecture=f16 Compression VAE2026.05 | 86.72 | 22.66 | 6.5 | 3.49 | 77.53 | |
| Wan2.2Setting=f16c48, Compression Factor=f16, Architecture=f16 Compression VAE2026.05 | 85.77 | 21.67 | 5.25 | 3.05 | 83.1 | |
| Qwen-Image-VAE-2.0-f32c128Setting=f32c128, Compression Factor=f32, Architecture=f32 Compression VAE2026.05 | 84.42 | 22.13 | 6.42 | 3.36 | 70.65 | |
| HunyuanVideo-1.5Setting=f16c32, Compression Factor=f16, Architecture=f16 Compression VAE2026.05 | 84.22 | 21.49 | 8.39 | 4.67 | 69.38 | |
| Wan2.1Setting=f8c16, Compression Factor=f8, Architecture=f8 Compression VAE2026.05 | 82.82 | 20.54 | 6.79 | 4.57 | 80.21 | |
| LTX-VideoSetting=f32c128, Compression Factor=f32, Architecture=f32 Compression VAE2026.05 | 80.55 | 20.92 | 11.9 | 17.1 | 56.51 | |
| HunyuanImage-2.1Setting=f32c64, Compression Factor=f32, Architecture=f32 Compression VAE2026.05 | 78.05 | 19.85 | 9.57 | 5.19 | 48.95 | |
| LTX-2Setting=f32c128, Compression Factor=f32, Architecture=f32 Compression VAE2026.05 | 73.54 | 18.41 | 11.92 | 9.94 | 35.69 | |
| VTP-LargeSetting=f16c64, Compression Factor=f16, Architecture=ViT-backbone AutoEncoder2026.05 | 71.85 | 18.11 | 10.46 | 3.94 | 41.7 | |
| VAVAESetting=f16c32, Compression Factor=f16, Architecture=f16 Compression VAE2026.05 | 69.05 | 17.5 | 9.74 | 4.45 | 34.88 | |
| Cosmos-0.1-CI16x16Setting=f16c16, Compression Factor=f16, Architecture=f16 Compression VAE2026.05 | 54.6 | 15.55 | 13.49 | 7.78 | 15.47 | |
| DC-AE-sanaSetting=f32c32, Compression Factor=f32, Architecture=f32 Compression VAE2026.05 | 52.59 | 15.62 | 14.41 | 7.26 | 6.92 | |
| RAE-SigLIP2-BSetting=f16c768, Compression Factor=f16, Architecture=ViT-backbone AutoEncoder2026.05 | 38.71 | 14.36 | 19.72 | 11.49 | 4.83 | |
| RAE-DINOv2-BSetting=f16c768, Compression Factor=f16, Architecture=ViT-backbone AutoEncoder2026.05 | 32.61 | 14.32 | 22.9 | 18.21 | 3.92 |