GLM-5.2 is a recent open-weight model release from Z.ai. My first impression is that it is the best open-weight model today. As usual for fresh releases, I would treat the release-time leaderboard position as date-sensitive.GLM-5.2 是 Z.ai 近期发布的一个开放权重模型。我的第一印象是，它是目前最好的开放权重模型。和往常一样，对于新发布的模型，我会将发布时的排行榜位置视为具有时效性。

Architecture-wise, it builds on the earlier GLM-5 and GLM-5.1 architecture. In particular, it reuses Multi-head Latent Attention and DeepSeek Sparse Attention, the DSA mechanism from DeepSeek V3.2 that I covered in the DeepSeek V3 to V3.2 article.在架构方面，它基于早期的 GLM-5 和 GLM-5.1 架构。具体来说，它复用了多头潜在注意力和 DeepSeek 稀疏注意力，即我在《DeepSeek V3 到 V3.2》一文中介绍的 DeepSeek V3.2 的 DSA 机制。

What’s new is IndexShare. This is a cross-layer reuse trick for DSA. Instead of recomputing the sparse-attention top-k indexer in every layer, GLM-5.2 runs the full indexer only once every four layers. The following layers then reuse the selected token indices.新特性是 IndexShare。这是 DSA 的一种跨层复用技巧。GLM-5.2 不是每层都重新计算稀疏注意力 top-k 索引器，而是每四层运行一次完整的索引器，后续层则复用选定的 token 索引。

This keeps the same DSA idea but makes 1M-token inference cheaper. The attention pattern is still adaptive, but the model spends less work repeatedly deciding which earlier tokens to attend to.这保留了相同的 DSA 思想，但降低了 100 万 token 推理的成本。注意力模式仍然是自适应的，但模型减少了重复决定关注哪些早期 token 的工作量。

The local GLM-5.2 architecture card has the current summary, config links, and benchmark references.本地 GLM-5.2 架构卡包含当前摘要、配置链接和基准测试参考。

GLM-5.2 architecture and benchmark overview

Composite figure from the original Substack note, summarizing the GLM-5.2 architecture and release-time average benchmark snapshot.来自原始 Substack 笔记的合成图，总结了 GLM-5.2 架构和发布时的平均基准测试快照。

By the way, in the Artificial Analysis Coding Index snapshot below, GLM-5.2 scores 68.8 versus 56.7 for Claude Opus 4.8 (max), more than 10 points higher on coding benchmarks. That’s super impressive for a fresh open-weight model.顺便提一下，在下面的人工分析编码指数快照中，GLM-5.2 得分为 68.8，而 Claude Opus 4.8（最大）为 56.7，在编码基准测试上高出 10 分以上。对于一个全新的开放权重模型来说，这非常令人印象深刻。

Artificial Analysis Coding Index chart comparing GLM-5.2 and Claude Opus 4.8

Artificial Analysis Coding Index snapshot showing GLM-5.2 at 68.8 and Claude Opus 4.8 (max) at 56.7.人工分析编码指数快照显示 GLM-5.2 为 68.8，Claude Opus 4.8（最大）为 56.7。

Source: lightly edited website version of my Substack note.来源：我的 Substack 笔记的网站版本，经过轻微编辑。