Event ·Official first-hand

Z.ai发布GLM-5.3-Flash:320B-A18B原生多模态MoE,1M令牌上下文窗口

Z.ai推出GLM-5.3-Flash,GLM-5系列首个原生多模态模型,320B总参数/18B激活MoE,1,048,576令牌上下文窗口,MIT授权Hugging Face权重,API定价0.15美元/输入和0.50美元/输出。在Terminal-Bench 2.1上得分84.3,在DeepSWE v1.1上得分63.4,使用混合KDA线性加NoPE稀疏MLA注意力减少注意力计算约3倍,KV缓存4.4倍于GLM-5.3。

Current read

Z.ai发布GLM-5.3-Flash,原生多模态MoE,上下文窗口更大,性价比高,值得尝试。

11 sources58° AI heatLast updated 2026/8/26 21:33:23

Evidence

11 sources
Primary sourcemarktechpost
View source
Related source 1Paul Couvert
View source
Related source 3lmarena.ai
View source
Related source 5shao__meng
View source
Related source 6OpenRouter
View source
Related source 7The Rundown AI
View source
Related source 9SuperTechFans
View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.