Gugu8 发布 L-CED-15B 循环 MoE 智能体编程模型
原标题：Gugu8/L-CED-15B
AI 摘要
Hugging Face 上出现名为 Gugu8/L-CED-15B 的新模型，宣称是一种面向智能体编程（agentic coding）的循环式因果编码器-解码器 MoE 架构。该模型总参数 15B、激活参数约 2-3.5B，支持 1M 上下文，并声称融合了 DeepSeek V4.1-Flash CED、CSA2、Engram、Single-Pass mHC、DSpark 等技术以及循环解码器和 Mixture-of-Recursions 路由。其目标是在 SWE-bench、Terminal-Bench、DeepSWE 上实现按激活参数计算的前沿推理能力。
正文节选
# L-CED 15B — Looped Causal Encoder-Decoder MoE for Agentic Coding Novel smartness-first architecture combining: - DeepSeek V4.1-Flash CED + CSA2 + Engram + Single-Pass mHC + DSpark - Looped decoder (recurrent depth) + Mixture-of-Recursions routing - 15B total, ~2-3.5B active, 1M context staged See ARCHITECTURE.md for full spec. Target: frontier reasoning per active param on SWE-bench / Terminal-Bench / DeepSWE. # IF YOU ARE USING THIS MODEL PRETTY PLEASE FOLLOW AND LIKE THIS REPO, NEW MODEL