Seed Diffusion: A Large-Scale Diffusion Language Model with High-Speed Inference

Authors:
  • YSYuxuan Song
    Yuxuan Song
  • CLCheng Luo
    Cheng Luo
  • PGPengyang Gao
    Pengyang Gao
  • FXFan Xia
    Fan Xia
  • HLHao Luo
    Hao Luo
  • ZLZheng Li
    Zheng Li
  • YYYuehang Yang
    Yuehang Yang
  • HYHongli Yu
    Hongli Yu
  • XQXingwei Qu
    Xingwei Qu
  • YFYuwei Fu
    Yuwei Fu
  • JSJing Su
    Jing Su

Abstract
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.
All Related Versions