Seed Diffusion: A Large-Scale Diffusion Language Model with High-Speed Inference
Authors:
- Yuxuan SongYSYuxuan Song
- Zheng ZhangZZZheng Zhang
- Cheng LuoCLCheng Luo
- Pengyang GaoPGPengyang Gao
- Fan XiaFXFan Xia
- Hao LuoHLHao Luo
- Zheng LiZLZheng Li
- Yuehang YangYYYuehang Yang
- Hongli YuHYHongli Yu
- Xingwei QuXQXingwei Qu
- Yuwei FuYFYuwei Fu
- Jing SuJSJing Su
Preprint
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.Keywords
All Related Versions
This publication has 3 references indexed in Scilit: