Language Model Beats Diffusion - Tokenizer is key to visual generation

Lijun Yu, José Lezama, N. B. Gundavarapu, Luca Versari, Kihyuk Sohn +10 more
2/12/2026

No official code found. Best implementation: kyegomez/MAGVIT2 (15 stars, MIT, last commit 2024-11-11).

Semantic Scholar

Code Implementations(1)

kyegomez/MAGVIT2Community67%
MIT

Open source community's implementation of the model from "LANGUAGE MODEL BEATS DIFFUSION — TOKENIZER IS KEY TO VISUAL GENERATION"

151PythonOct 11, 20231 years agoMIT

Cite this paper

@article{yu2026language,
  title  = {Language Model Beats Diffusion - Tokenizer is key to visual generation},
  author = {Lijun Yu and José Lezama and N. B. Gundavarapu and Luca Versari and Kihyuk Sohn and David Minnen and Yong Cheng and Agrim Gupta and Xiuye Gu and Alexander G. Hauptmann and Boqing Gong and Ming-Hsuan Yang and Irfan Essa and David A. Ross and Lu Jiang},
  year   = {2026},
  url    = {https://api.semanticscholar.org/CorpusID:aaa1844f05acc7c510287a9df811f9324fcf44fb},
  journal = {ICLR 2024 2024}
}

Discussion