智谱 AI · Z.aiZ.ai (Zhipu AI)　·　当前代current generation GLM-5.3GLM-5.3（2026-08）

GLM 智谱GLM (Z.ai)模型家族谱系卡model family lineage card

中文 agent 与代码强项、迭代节奏极快（半年四代）。大杯与 Flash 杯许可不同——本地部署优先 MIT 的 Flash 杯。本站收录的 glm-5 系列在 GLM-5.3 上。Strong at Chinese agentic work and code, with a very fast iteration cadence. Flagship and Flash tiers carry different licences — for local deployment, prefer the MIT Flash tier. This site's glm-5 series tracks GLM-5.3.

本站已收录该家族的量化对比：This site covers this family's quantisations: GLM 5GLM 5

先说清楚：这个站的数据怎么来的First, where this site's data comes from

✅ 实测字段Measured fields

直接从 Hugging Face 公开 API 读取的客观数据：文件体积（字节）、上下文长度、架构、许可、下载量、更新时间。Objective data read straight from the Hugging Face public API: file sizes in bytes, context length, architecture, licence, downloads, last updated.

✅ 推导Derived

由实测字段按公开规则算出：每权重比特数（依 llama.cpp 命名标准）、档位归属、显存需求区间。推导规则在Computed from the measured fields by published rules: bits per weight (per the llama.cpp naming convention), tier assignment, VRAM requirement range. The rules are fully published in METHODOLOGY.md 完整公开，任何人可复算。 — anyone can recompute them.

❌ 我们不说的What we do not claim

哪个量化质量更高、哪个更快、哪个模型更强——这些需要实测，本站不跑 benchmark，所以不装作知道。凡我们给出的主观判断，一律显式标注「我们的口径」。Which quantisation is higher quality, which is faster, which model is stronger — these need measurement. This site does not run benchmarks, so it does not pretend to know. Every subjective judgement we do offer is explicitly marked as our own reading.

本站This site does 不下载、不托管任何模型权重文件not download or host any model weight files。所有数字都来自公开元数据，点击即可回到发布页核对。. Every figure comes from public metadata, and clicking it takes you back to the release page to verify. 采集时间Collected: 2026-10-30（数据快照，非实时） (a snapshot, not live)

中文生态China availability 国产 / 国内可直接用 / 备案状态 · 快照domestic / direct access / CAC filing · snapshot 2026-10-20

出品属地Origin　国产domestic　出品方为智谱 AI（北京）。Made by Zhipu AI (Beijing).
国内下载Direct access　国内可直连direct OK　ModelScope 有 ZhipuAI（「智谱.AI」）官方组织镜像（2026-10-20 实测 200）；Hugging Face 直连不可达。Official ZhipuAI org mirror on ModelScope (measured 200); Hugging Face direct unreachable.　·　国内镜像入口domestic mirror
备案状态CAC filing　已备案Filed：Beijing-ChatGLM-20230821　公示名称listed as「智谱清言（ChatGLM）」　·　备案时间filed 2023-08-31　·　首批第 2 行。Row 2 of the first batch.

备案信息以网信办《生成式人工智能服务已备案信息》公示名单为准（按批次滚动更新，本站只抄编号不解读）；「国内可直连」是本站从大陆网络直连实测的结果，不是厂商承诺。Filing status is quoted from the CAC public registry (updated batch by batch; this site copies the number and does not interpret it). "Direct access" is this site's own reachability measurement from a mainland network, not a vendor promise.

网信办《生成式人工智能服务已备案信息》滚动公告页（13 个批次附件 PDF，2026-10-20 全量下载检索）：CAC · filed generative-AI service information (rolling announcement page, 13 batch attachment PDFs, full download & search on 2026-10-20): https://www.cac.gov.cn/2024-04/02/c_1713729983803145.htm
ModelScope 官方组织镜像（2026-10-20 实测 200）：ModelScope official-org mirror (200 verified on 2026-10-20): https://www.modelscope.cn/models/ZhipuAI/GLM-5.3

代次Generations

GLM-5.3GLM-5.32026-08当前代current

当前代。754B MoE（激活数未公开）+ 321B Flash 杯；官方口径「open-weights 里编码最强」。Current generation. 754B MoE (activation not disclosed) plus a 321B Flash tier; officially billed as the strongest open-weights coder.
官方页面 ↗official page ↗
GLM-5.2 / 5.1GLM-5.2 / 5.12026-06

同代中期两次迭代；5.2 的代码基准是 5.3 官方改进幅度的对照（「50% improvement over GLM-5.2」）。Two mid-generation refreshes; GLM-5.2 is the reference point for the official 5.3 improvement claim ("50% improvement over GLM-5.2").
官方页面 ↗official page ↗
GLM-5GLM-52026-02

本代起点，同月发布 FP8 变体。Generation opener, with an FP8 variant released the same month.
官方页面 ↗official page ↗
GLM-4.7-FlashGLM-4.7-Flash2026-01

上一代 Flash 杯，MIT 许可的小杯路线。Previous-generation Flash tier, continuing the small MIT-licensed line.
官方页面 ↗official page ↗

尺寸档Size tiers

小杯 Flash：MIT 许可，本地部署优先选它Flash tier: MIT-licensed, the first choice for local deployment
大杯 5.3：754B MoE，服务器级；许可以为发布页原文为准Flagship 5.3: 754B MoE, server-class; licence per the release page

许可Licences

mitMIT（GLM-5.3-Flash 等 Flash 杯）MIT (GLM-5.3-Flash and other Flash tiers) 许可原文 ↗licence text ↗
otherother · license_name: glm-5.3（大杯，以发布页原文为准）other · license_name: glm-5.3 (flagship; see the release page) 许可原文 ↗licence text ↗

官方能力声明（带出处）Official capability claims (with sources)

官方 README 自评：「GLM-5.3 is the most capable open-weights model for coding, with a 50% improvement over GLM-5.2 on our in-house Z.ai Code Bench」——「本站口径」风险点：这是厂商自评基准，不是第三方横评。Official README self-assessment: "GLM-5.3 is the most capable open-weights model for coding, with a 50% improvement over GLM-5.2 on our in-house Z.ai Code Bench" — note this is a vendor-run benchmark, not third-party evaluation.出处 ↗source ↗
官方 config：max_position_embeddings = 1,048,576（1M 上下文）；官方 README 的 NL2Repo 评测在 1M 上下文下进行。Official config: max_position_embeddings = 1,048,576 (1M context); the README's NL2Repo evaluation ran under 1M context.出处 ↗source ↗

以上均为厂商官方页面原文或官方自评基准的转述；涉及横评结论的，本站不背书——本站不跑 benchmark。The above are quotations of official pages or transcriptions of vendor-run benchmarks; this site does not endorse cross-model conclusions — it runs no benchmarks.

出处Sources

GLM-5.3 模型页与 README（2026-10-20 直抓 200）：GLM-5.3 model page & README (direct fetch 200 on 2026-10-20): https://huggingface.co/zai-org/GLM-5.3
GLM-5.3-Flash 模型页（许可核验，2026-10-20 直抓 200）：GLM-5.3-Flash model page (licence check, direct fetch 200 on 2026-10-20): https://huggingface.co/zai-org/GLM-5.3-Flash
GLM-5.2 模型页（代次核验，2026-10-20 直抓 200）：GLM-5.2 model page (generation check, direct fetch 200 on 2026-10-20): https://huggingface.co/zai-org/GLM-5.2
GLM-4.7-Flash 模型页（代次核验，2026-10-20 直抓 200）：GLM-4.7-Flash model page (generation check, direct fetch 200 on 2026-10-20): https://huggingface.co/zai-org/GLM-4.7-Flash

核验快照：2026-10-30 · 本站不实测、不排名，未知就写未知。Verification snapshot: 2026-10-30 · no benchmarks, no rankings — unknown stays unknown.