Buzzing HN · 中文精选· pretext·· 2 hours agoSelectedAI score80
Baidu Releases ERNIE 5.5 Flash Base Foundation Model
百度发布了ERNIE 5.5 Flash基础模型
AI guide
Baidu has officially released its foundation model, ERNIE-5.5-Flash-Base, featuring a sparse MoE architecture with 280B total parameters and 14B activated per token, pre-trained from scratch on 33.5T tokens.
Why it matters
The paper discloses full architectural details of hybrid sparse attention and multi-routing MoE alongside training benchmarks, helping developers evaluate computational overhead when using it as a base for downstream training.
Source:Buzzing HN · 中文精选 · ernie.baidu.com