| WeTok: Powerful Discrete Tokenization for High-Fidelity Visual Reconstruction | ✓ Link | 0.19 | | | | WeTok (32x32) | 2025-08-07 |
| HieraTok: Multi-Scale Visual Tokenizer Improves Image Reconstruction and Generation | | 0.45 | | | | HieraTok | 2025-09-28 |
| MGVQ: Could VQ-VAE Beat VAE? A Generalizable Tokenizer with Multi-group Quantization | ✓ Link | 0.49 | 0.086 | 24.70 | 0.787 | MGVQ (16x16x8) | 2025-07-14 |
| 2D Gaussians Meet Visual Tokenizer | | 0.556 | | | | GQT-multigs | 2025-08-19 |
| WeTok: Powerful Discrete Tokenization for High-Fidelity Visual Reconstruction | ✓ Link | 0.61 | | | | WeTok (16x16) | 2025-08-07 |
| MGVQ: Could VQ-VAE Beat VAE? A Generalizable Tokenizer with Multi-group Quantization | ✓ Link | 0.64 | 0.110 | 23.71 | 0.755 | MGVQ (16x16x4) | 2025-07-14 |
| Balancing Image Compression and Generation with Bootstrapped Tokenization | | 0.66 | | | | SelfBootTok | 2026-06-04 |
| GigaTok: Scaling Visual Tokenizers to 3 Billion Parameters for Autoregressive Image Generation | ✓ Link | 0.79 | 0.1947 | 21.65 | 0.699 | GigaTok-XL-XXL | 2025-04-11 |
| UniWeTok: An Unified Binary Tokenizer with Codebook Size $\mathit{2^{128}}$ for Unified Multimodal Large Language Model | ✓ Link | 0.79 | | | | UniWeTok | 2026-02-15 |
| Channel-wise Vector Quantization | ✓ Link | 0.88 | | | | CVQ (1024 tokens) | 2026-05-25 |
| Vision Foundation Models as Effective Visual Tokenizers for Autoregressive Image Generation | ✓ Link | 0.89 | | | | VFMTok | 2025-07-11 |
| Preventing Local Pitfalls in Vector Quantization via Optimal Transport | ✓ Link | 0.91 | 0.066 | 27.57 | 0.729 | OptVQ (16x16x8) | 2024-12-19 |
| Taming Scalable Visual Tokenizer for Autoregressive Image Generation | ✓ Link | 1.00 | 0.2030 | | | IBQ (16x16) | 2024-12-03 |
| Preventing Local Pitfalls in Vector Quantization via Optimal Transport | ✓ Link | 1.00 | 0.076 | 26.59 | 0.717 | OptVQ (16x16x4) | 2024-12-19 |
| 2D Gaussians Meet Visual Tokenizer | | 1.00 | | | | GQT | 2025-08-19 |
| LooC: Effective Low-Dimensional Codebook for Compositional Vector Quantization | | 1.01 | | | | LooC-VAE (1024x4) | 2026-01-01 |
| Improving Flexible Image Tokenizers for Autoregressive Image Generation | ✓ Link | 1.01 | | | | ReTok-S-B | 2026-01-04 |
| Improving Flexible Image Tokenizers for Autoregressive Image Generation | ✓ Link | 1.09 | | | | ReTok-S-S | 2026-01-04 |
| MoVQ: Modulating Quantized Vectors for High-Fidelity Image Generation | ✓ Link | 1.12 | 0.113 | 22.42 | 0.673 | Mo-VQGAN (16x16x4) | 2022-09-19 |
| Open-MAGVIT2: An Open-Source Project Toward Democratizing Auto-regressive Visual Generation | ✓ Link | 1.17 | | 21.90 | | Open-Magvit2 (16x16) | 2024-09-06 |
| Vector-quantized Image Modeling with Improved VQGAN | ✓ Link | 1.28 | | | | ViT-VQGAN (16x16) | 2021-10-09 |
| MaskBit: Embedding-free Image Generation via Bit Tokens | ✓ Link | 1.66 | | | | MaskBit (16x16) | 2024-09-24 |
| LooC: Effective Low-Dimensional Codebook for Compositional Vector Quantization | | 1.68 | | | | LooC-VAE (256x4) | 2026-01-01 |
| An Image is Worth 32 Tokens for Reconstruction and Generation | ✓ Link | 1.71 | | | | TiTok-S-128 | 2024-06-11 |
| Autoregressive Image Generation using Residual Quantization | ✓ Link | 1.83 | | | | RQ-VAE (8x8x16) | 2022-03-03 |
| MaskGIT: Masked Generative Image Transformer | ✓ Link | 2.28 | | | | MaskGIT-VQGAN (16x16) | 2022-02-08 |
| Scaling the Codebook Size of VQGAN to 100,000 with a Utilization Rate of 99% | ✓ Link | 2.62 | 0.120 | 23.80 | 0.589 | VQGAN-LC (16x16) | 2024-06-17 |
| Taming Transformers for High-Resolution Image Synthesis | ✓ Link | 3.64 | 0.177 | 19.93 | 0.542 | Taming-VQGAN (16x16) | 2020-12-17 |