NextStep-1: Toward Autoregressive Image Generation with Continuous Tokens at Scale
Abstract
Prevailing autoregressive (AR) models for text-to-image generation either rely on heavy, computationally-intensive diffusion models to process continuous image tokens, or employ vector quantization (VQ) to obtain discrete tokens with quantization loss. In this paper, we push the autoregressive paradigm forward with NextStep-1, a 14B autoregressive model paired with a 157M flow matching head, trained on discrete text tokens and continuous image tokens with next-token prediction objectives. NextStep-1 achieves state-of-the-art performance for autoregressive models in text-to-image generation tasks, exhibiting strong capabilities in high-fidelity image synthesis. Furthermore, our method shows strong performance in image editing, highlighting the power and versatility of our unified approach. To facilitate open research, we have released our code and models to the community at https://github.com/stepfun-ai/NextStep-1.
Cite
Text
Han et al. "NextStep-1: Toward Autoregressive Image Generation with Continuous Tokens at Scale." International Conference on Learning Representations, 2026.Markdown
[Han et al. "NextStep-1: Toward Autoregressive Image Generation with Continuous Tokens at Scale." International Conference on Learning Representations, 2026.](https://mlanthology.org/iclr/2026/han2026iclr-nextstep1/)BibTeX
@inproceedings{han2026iclr-nextstep1,
title = {{NextStep-1: Toward Autoregressive Image Generation with Continuous Tokens at Scale}},
author = {Han, Chunrui and Li, Guopeng and Wu, Jingwei and Sun, Quan and Cai, Yan and Peng, Yuang and Ge, Zheng and Zhou, Deyu and Tang, Haomiao and Zhou, Hongyu and Liu, Kenkun and Xia, Shu-Tao and Jiao, Binxing and Jiang, Daxin and Zhang, Xiangyu and Zhu, Yibo},
booktitle = {International Conference on Learning Representations},
year = {2026},
url = {https://mlanthology.org/iclr/2026/han2026iclr-nextstep1/}
}