Published at 11/5/2026, 12:10:43 PM · 26 views
📰 Baidu Opens ERNIE-5.5-Flash-Base: 280B-A14B MoE, 33.5T Pretrain, Vendor-Reported MMLU 93.3
On November 5 Baidu (official blog + @BaiduResearch) open-sourced ERNIE-5.5-Flash-Base: a text-only MoE base checkpoint at end of pre-training (no mid/post-training), 280B total / 14B active per token, 33.5T pretrain tokens, up to 128K train context and 1M for fine-tune/inference.
💬 What's your take? Join the discussion in the comments.
Read the original Chinese version
📰 百度开源 ERNIE-5.5-Flash-Base：280B 总参 / 14B 激活 MoE，预训练 33.5T，厂商口径 MMLU 93.3
百度 11 月 5 日通过官网与 @BaiduResearch 宣布开源 ERNIE-5.5-Flash-Base：预训练结束点、未经中段/后训练的纯文本 MoE 基座，总参 280B、每 token 激活 14B，预训练 33.5T、训练上下文最长 128K，微调与推理支持 1M。
💬 你怎么看？欢迎在评论区讨论。
#ERNIE 5.5 #Baidu #MoE
0 Comments