Event ·Official first-hand

GLM-5.3-Flash与Qwen3.8-Flash-Next架构趋同

Z.ai和Qwen两家中国AI实验室独立发布了几乎相同的模型架构。GLM-5.3-Flash和Qwen3.8-Flash-Next都采用了3:1线性混合架构。两家模型都使用了压缩索引器和门控残差设计。它们还都采用了Muon训练方法。

Current read

两家中国实验室不约而同开发出相似架构，看看GLM和Qwen的最新flash版本有何异同。

10 sources73° AI heatLast updated 2026/8/28 19:12:21

Evidence

10 sources

Primary sourcemarktechpost

2026/8/28 19:12:21View source

Related source 1SuperTechFans

2026/8/27 00:44:28View source

Related source 2阿里通义 Qwen

2026/8/27 03:00:36View source

Related source 3AI Will

2026/8/28 03:39:25View source

Related source 4pandaily

2026/8/27 02:40:53View source

Related source 5Geek

2026/8/27 04:30:28View source

Related source 6Decoder

2026/8/27 10:24:49View source

Related source 7Fireworks AI

2026/8/29 06:50:32View source

Related source 8歸藏(guizang.ai)

2026/8/28 07:19:34View source

Related source 9ollama

2026/8/29 00:13:31View 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.

Open the story