趋近智
GLM-5
活跃参数
744B
上下文长度
204.8K
模态
Multimodal
架构
Mixture of Experts (MoE)
许可证
MIT
发布日期
12 Feb 2026
训练数据截止日期
-
技术规格
专家参数总数
40.0B
专家数量
-
活跃专家
-
注意力结构
Multi-Head Attention
隐藏维度大小
-
层数
-
注意力头
-
键值头
-
激活函数
-
归一化
-
位置嵌入
Absolute Position Embedding
系统要求
不同量化方法和上下文大小的显存要求
GLM-5
GLM-5 is a state-of-the-art multimodal model from Z.ai (Zhipu AI) released in February 2026. Built on a 744 billion parameter Mixture-of-Experts (MoE) architecture with 40 billion active parameters, it integrates DeepSeek Sparse Attention (DSA) to deliver massive intelligence efficiency while maintaining long-context capacity (200K+ tokens). Released under the MIT License, GLM-5 targets complex systems engineering and long-horizon agentic tasks, achieving top-tier performance on SWE-bench and Vending Bench 2.
关于 GLM 5
GLM 5 is the fifth generation of General Language Models developed by Z.ai. It represents a significant leap in multimodal foundational capabilities, featuring advanced reasoning and long-horizon agentic capabilities across diverse systems engineering tasks.
其他 GLM 5 模型
- 没有相关模型
评估基准
没有可用的 GLM-5 评估基准。