Saved in:
Bibliographic Details
Main Authors: Song, Yuxuan, Zhang, Zheng, Luo, Cheng, Gao, Pengyang, Xia, Fan, Luo, Hao, Li, Zheng, Yang, Yuehang, Yu, Hongli, Qu, Xingwei, Fu, Yuwei, Su, Jing, Zhang, Ge, Huang, Wenhao, Wang, Mingxuan, Yan, Lin, Jia, Xiaoying, Liu, Jingjing, Ma, Wei-Ying, Zhang, Ya-Qin, Wu, Yonghui, Zhou, Hao
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2508.02193
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908477846192128
author Song, Yuxuan
Zhang, Zheng
Luo, Cheng
Gao, Pengyang
Xia, Fan
Luo, Hao
Li, Zheng
Yang, Yuehang
Yu, Hongli
Qu, Xingwei
Fu, Yuwei
Su, Jing
Zhang, Ge
Huang, Wenhao
Wang, Mingxuan
Yan, Lin
Jia, Xiaoying
Liu, Jingjing
Ma, Wei-Ying
Zhang, Ya-Qin
Wu, Yonghui
Zhou, Hao
author_facet Song, Yuxuan
Zhang, Zheng
Luo, Cheng
Gao, Pengyang
Xia, Fan
Luo, Hao
Li, Zheng
Yang, Yuehang
Yu, Hongli
Qu, Xingwei
Fu, Yuwei
Su, Jing
Zhang, Ge
Huang, Wenhao
Wang, Mingxuan
Yan, Lin
Jia, Xiaoying
Liu, Jingjing
Ma, Wei-Ying
Zhang, Ya-Qin
Wu, Yonghui
Zhou, Hao
contents We present Seed Diffusion Preview, a large-scale language model based on discrete-state diffusion, offering remarkably fast inference speed. Thanks to non-sequential, parallel generation, discrete diffusion models provide a notable speedup to mitigate the inherent latency of token-by-token decoding, as demonstrated recently (e.g., Mercury Coder, Gemini Diffusion). Seed Diffusion Preview achieves an inference speed of 2,146 token/s over H20 GPUs while maintaining competitive performance across a sweep of standard code evaluation benchmarks, significantly faster than contemporary Mercury and Gemini Diffusion, establishing new state of the art on the speed-quality Pareto frontier for code models.
format Preprint
id arxiv_https___arxiv_org_abs_2508_02193
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Seed Diffusion: A Large-Scale Diffusion Language Model with High-Speed Inference
Song, Yuxuan
Zhang, Zheng
Luo, Cheng
Gao, Pengyang
Xia, Fan
Luo, Hao
Li, Zheng
Yang, Yuehang
Yu, Hongli
Qu, Xingwei
Fu, Yuwei
Su, Jing
Zhang, Ge
Huang, Wenhao
Wang, Mingxuan
Yan, Lin
Jia, Xiaoying
Liu, Jingjing
Ma, Wei-Ying
Zhang, Ya-Qin
Wu, Yonghui
Zhou, Hao
Computation and Language
Machine Learning
We present Seed Diffusion Preview, a large-scale language model based on discrete-state diffusion, offering remarkably fast inference speed. Thanks to non-sequential, parallel generation, discrete diffusion models provide a notable speedup to mitigate the inherent latency of token-by-token decoding, as demonstrated recently (e.g., Mercury Coder, Gemini Diffusion). Seed Diffusion Preview achieves an inference speed of 2,146 token/s over H20 GPUs while maintaining competitive performance across a sweep of standard code evaluation benchmarks, significantly faster than contemporary Mercury and Gemini Diffusion, establishing new state of the art on the speed-quality Pareto frontier for code models.
title Seed Diffusion: A Large-Scale Diffusion Language Model with High-Speed Inference
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2508.02193