Z.ai 發表 GLM-5.2:具備可用 100 萬 Token 上下文、兩種思考努力層級,發表時未附基準測試
重點摘要
GLM-5.2 是 Z.ai 最新的大型語言模型,也是 GLM-5 系列的第三個主要版本。繼 GLM-5(2 月 11 日)、GLM-5-Turbo(3 月 15 日)和 GLM-5.1(4 月 7 日)之後,這已是約四個月內第四個旗艦級編碼用版本。其最顯著的規格是 100 萬 Token 的上下文視窗,Z.ai 在自身配置中將此變體標記為 glm-5.2[1m]。每次回應最多可輸出 131,072 個 Token,比 GLM-5.1 的 20 萬 Token 視窗提升了約 5 倍。百萬級 Token 視窗改變了編碼代理的實際運作方式:代理能將整個中型程式碼庫(包括原始碼、測試檔、設定檔與對話歷史)保留在工作記憶中,避免了較小視窗強制進行的頻繁摘要動作。
GLM-5.2 is the latest large language model from Z.ai, becoming the third major release in the GLM-5 line. It follows GLM-5 (February 11), GLM-5-Turbo (March 15), and GLM-5.1 (April 7). That makes four flagship-tier coding releases in roughly four months. Usable 1M-Token Context Window GLM-5.2’s standout spec is a 1,000,000-token context window. Z.ai labels the variant glm-5.2[1m] in its own configuration. Each response can return up to 131,072 output tokens. That is roughly a 5x jump from GLM-5.1’s 200,000-token window. A 1M-token window changes how a coding agent works in practice. The agent can hold an entire mid-sized repository in working memory. That includes source files, tests, configuration, and conversation history. It avoids the constant summarization that smaller windows force. The release also adds two thinking-effort levels: High and Max. Z.ai recommends Max effort for complex, multi-step coding work. In Claude Code, the /effort command controls this setting. The xhigh, max, and ultracode options all map to GLM-5.2’s Max effort. Architecture and What Changed Z.ai did not specify GLM-5.2’s architecture in its launch materials. But based on community notes, the GLM-5 base is a 744-billion-parameter Mixture-of-Experts model. It activates 40 billion parameters per token. GLM-5.1 kept that same backbone with retargeted post-training. MTP Explainer Playground #mtp-glm52-demo, #mtp-glm52-demo * { box-sizing:border-box!important; } #mtp-glm52-demo hr, #mtp-glm52-demo p:empty, #mtp-glm52-demo del, #mtp-glm52-demo s { display:none!important; } #mtp-glm52-demo { background:#111!important; color:#e8e8e8!important; border:1px solid #2a2a2a!important; border-radius:14px!important; font-family:-apple-system,BlinkMacSystemFont,"Segoe UI",Roboto,Helvetica,Arial,sans-serif!important; max-width:900px!important; margin:24px auto!important; padding:24px!important; line-height:1.5!important; } #mtp-glm52-demo .g52-head { border-bottom:1px solid #2a2a2a!important; padding-bottom:14px!important; margin-bottom:18px!important; } #mtp-glm52-demo .g52-kicker { color:#76B900!important; font-size:12px!important; letter-spacing:.12em!important; text-transform:uppercase!important; font-weight:700!important; margin:0 0 6px!important; } #mtp-glm52-demo .g52-title { font-size:21px!important; font-weight:700!important; color:#fff!important; margin:0!important; } #mtp-glm52-demo .g52-sub { color:#9a9a9a!important; font-size:13px!important; margin:6px 0 0!important; } #mtp-glm52-demo .g52-label { font-size:11px!important; letter-spacing:.08em!important; text-transform:uppercase!important; color:#888!important; font-weight:700!important; margin:0 0 8px!important; } #mtp-glm52-demo .g52-row { margin-bottom:18px!important; } #mtp-glm52-demo .g52-btns { display:flex!important; flex-wrap:wrap!important; gap:8px!important; } #mtp-glm52-demo .g52-btn { background:#1a1a1a!important; color:#cfcfcf!important; border:1px solid #333!important; border-radius:8px!important; padding:9px 14px!important; font-size:13px!important; font-weight:600!important; cursor:pointer!important; transition:all .15s ease!important; } #mtp-glm52-demo .g52-btn:hover { border-color:#76B900!important; color:#fff!important; } #mtp-glm52-demo .g52-btn.is-on { background:#76B900!important; color:#0a0a0a!important; border-color:#76B900!important; } #mtp-glm52-demo .g52-bars { margin:4px 0 0!important; } #mtp-glm52-demo .g52-bar-wrap { margin-bottom:12px!important; } #mtp-glm52-demo .g52-bar-top { display:flex!important; justify-content:space-between!important; font-size:12px!important; color:#bbb!important; margin-bottom:5px!important; } #mtp-glm52-demo .g52-bar-top b { color:#fff!important; } #mtp-glm52-demo .g52-track { background:#1d1d1d!important; border-radius:6px!important; height:16px!important; overflow:hidden!important; border:1px solid #2a2a2a!important; } #mtp-glm52-demo .g52-fill { height:100%!important; border-radius:6px 0 0 6px!important; transition:width .6s cubic-bezier(.2,.7,.3,1)!important; } #mtp-glm52-demo .g52-fill.old { background:#3d4d1a!important; } #mtp-glm52-demo .g52-fill.new { background:#76B900!important; } #mtp-glm52-demo .g52-code-wrap { position:relative!important; } #mtp-glm52-demo pre { background:#0c0c0c!important; color:#dcdcdc!important; border:1px solid #2a2a2a!important; border-radius:10px!important; padding:16px!important; margin:0!important; font-family:"SFMono-Regular",Consolas,"Liberation Mono",Menlo,monospace!important; font-size:12.5px!important; line-height:1.6!important; overflow-x:auto!important; white-space:pre!important; } #mtp-glm52-demo pre code { color:#dcdcdc!important; background:none!important; padding:0!important; } #mtp-glm52-demo .g52-k { color:#76B900!important; } #mtp-glm52-demo .g52-s { color:#c5a3ff!important; } #mtp-glm52-demo .g52-copy { position:absolute!important; top:10px!important; right:10px!important; background:#1a1a1a!important; color:#76B900!important; border:1px solid #76B900!important; border-radius:6px!important; padding:5px 11px!important; font-size:11px!important; font-weight:700!important; cursor:pointer!important; } #mtp-glm52-demo .g52-copy:hover { background:#76B900!important; color:#0a0a0a!important; } #mtp-glm52-demo .g52-note { font-size:12.5px!important; color:#9a9a9a!important; margin:12px 0 0!important; padding:11px 13px!important; background:#161616!important; border-left:3px solid #76B900!important; border-radius:0 8px 8px 0!important; } #mtp-glm52-demo .g52-note b { color:#cfcfcf!important; } #mtp-glm52-demo .g52-fits { display:flex!important; flex-wrap:wrap!important; gap:10px!important; margin-top:4px!important; } #mtp-glm52-demo .g52-chip { background:#161616!important; border:1px solid #2a2a2a!important; border-radius:8px!important; padding:10px 12px!important; flex:1!important; min-width:120px!important; } #mtp-glm52-demo .g52-chip .n { color:#76B900!important; font-size:18px!important; font-weight:700!important; display:block!important; } #mtp-glm52-demo .g52-chip .t { color:#9a9a9a!important; font-size:11px!important; } #mtp-glm52-demo .g52-foot { border-top:1px solid #2a2a2a!important; margin-top:20px!important; padding-top:12px!important; font-size:11.5px!important; color:#777!important; display:flex!important; justify-content:space-between!important; flex-wrap:wrap!important; gap:6px!important; } #mtp-glm52-demo .g52-foot b { color:#76B900!important; } @media (max-width:640px){ #mtp-glm52-demo { padding:16px!important; } #mtp-glm52-demo .g52-title { font-size:18px!important; } #mtp-glm52-demo .g52-btn { padding:8px 11px!important; font-size:12px!important; } #mtp-glm52-demo pre { font-size:11.5px!important; padding:13px!important; } #mtp-glm52-demo .g52-chip { min-width:calc(50% - 5px)!important; } } Interactive Demo GLM-5.2 Setup Generator & Context Visualizer Pick your agent and effort mode. Copy the exact config. See what 1M tokens buys you. 1. Coding agent Claude Code Claude Code (env vars) OpenClaw Cline 2. Context window 1M tokens (glm-5.2[1m]) Standard (glm-5.2) 3. Thinking effort Max (complex coding) High Your config Copy Context window: GLM-5.1 vs GLM-5.2 GLM-5.1~200,000 tokens GLM-5.21,000,000 tokens GLM-5.2 at a glance 1,000,000input tokens in one context window 131,072max output tokens per response 5xlarger than GLM-5.1’s window 8agentic tools supported day one Config sourced from Z.ai developer docs · June 2026 © Marktechpost (function(){ var root = document.getElementById('mtp-glm52-demo'); if(!root) return; var state = { tool:'claude', ctx:'1m', eff:'max' }; function model(){ return state.ctx==='1m' ? 'glm-5.2[1m]' : 'glm-5.2'; } function effortLine(){ return state.eff==='max' ? 'Run /effort in your session and select max for deeper reasoning.' : 'Run /effort in your session and select high for faster turns.'; } function configs(){ var m = model(); var compact = state.ctx==='1m' ? '1000000' : '200000'; if(state.tool==='claude'){ return { text:'{\n "env": {\n "CLAUDE_CODE_AUTO_COMPACT_WINDOW": "'+c
Related
相關文章
Liquid AI Introduces LFM2.5-Embedding-350M and LFM2.5-ColBERT-350M: Dense Bi-Encoder and Late-Interaction Models for Fast Multilingual Search Across 11 Languages
This week, Liquid AI released two new retrieval models. They are LFM2.5-ColBERT-350M and LFM2.5-Embedding-350M. Both hold 350M parameters. Both are the first bidirectional members of the LFM family. They build on LFM2.5-350M-Base, released in March. The pair targets fast multilingual and cross-lingual search across 11 languages. Their footprint is small enough to run almost anywhere. Both are available now on Hugging Face under the LFM Open License v1.0. LFM2.5 Retrievers The two models share one backbone but represent text differently. LFM2.5-Embedding-350M is a dense bi-encoder. It turns each document into a single vector. Pick it when you want the fastest search and the smallest, cheapest index. LFM2.5-ColBERT-350M is a late-interaction model. It converts each token into a vector rather
Perplexity Launches Brain, a Self-Improving Memory System That Builds a Context Graph of an Agent’s Work and Learns Overnight
Most AI memory remembers the user. It stores your preferences, your tastes, and your role. Perplexity is taking a different path. Today, Perplexity launched Brain, a self-improving memory system for its agent product, Computer. Brain does not focus on remembering you. It remembers what the agent did. That reframes what memory in AI is for. What is Perplexity‘s Brain Brain is a self-improving memory system. It builds a context graph of the work Computer performs. At set intervals, such as overnight, Brain reviews that graph. It then teaches itself how to do the work better. The idea is straightforward. The more work you do, the more efficient Brain makes your Computer. Brain is rolling out today to Perplexity Max and Enterprise Max subscribers in Research Preview. Two Axes of AI Memory Perp

智譜新高,MiniMax承壓,“大模型雙雄”命運殊途
這篇消息聚焦「智譜新高,MiniMax承壓,“大模型雙雄”命運殊途」。原始導語提到:大模型在被市場重新定價 從 AI 情報角度來看,這類內容值得關注其背後的技術進展、產品落地、產業競爭與後續市場影響。

華為昇騰 0 Day 支持智譜 GLM-5.2 模型,提供全面推理優化
華為昇騰 AI 宣佈在智譜開源 GLM-5.2 大模型當天即完成深度推理優化。通過 MOE 大融合算子、通信計算融合、高併發調度等七項關鍵技術,顯著提升編程和長程任務的處理效率,現已支持 A3 系列產品部署。#AI 大模型# #國產算力#
企業AI轉型再添利器:青雲科技算力雲接入 MiniMax-M3 模型
企業AI落地面臨高效低成本難題。青雲科技旗下基石智算平臺接入國產開源大模型MiniMax-M3,提供新算力支持。MiniMax-M3以卓越上下文處理能力等三大核心技術見長,依託自研架構,助企業便捷部署AI業務。
阿里開源統一科學大模型 LOGOS,僅用五十六分之一參數超越微軟
阿里 ATH-Token Foundry 聯閤中國人民大學高瓴人工智能學院開源科學基礎模型 LOGOS。該模型採用統一科學語法與純序列建模範式,在六大科學任務上匹配或超越傳統專用方法。其中 LOGOS-1B 僅 1B 參數,即展現出極高效率,性能超越參數量達 8×7B 的微軟模型。