Every Attention Matters: An Efficient Hybrid Architecture for Long-Context Reasoning

Ling Team, Bin Han, Caizhi Tang +25

In this technical report, we present the Ring-linear model series, specifically including Ring-mini-linear-2.0 and Ring-flash-linear-2.0. Ring-mini-linear-2.0 comprises 16B paramet…