电商热点雷达运营热点解读
meituan-longcat/LongCat-Flash-Lite-Sparse
雷达摘要
LongCat-Flash-Lite-Sparse is a non-thinking Mixture-of-Experts (MoE) model with 69B total parameters and approximately 3B activated parameters per token. Built on LongCat-Flash-Lite , it replaces dense MLA with LongCat Sparse Attention (LSA) and natively supports context lengths of up to 1M tokens. Compared with its dense predecessor, LongCat-Flash-Lite-Spa…
原始信息
本页是基于公开来源生成的摘要与运营提示，不替代原文。请以原始发布方内容为准。