★ Reading this for free? Get 20 structured AI courses + per-chapter AI tutor — the first chapter of every course free, no card.Start free in 30 seconds

Z.ai · Open Weights · MIT License

GLM-5.2: The Biggest Open Model You Can Actually Self-Host

GLM-5.2 (Z.ai, June 2026) is a 753B-parameter Mixture-of-Experts model — roughly 40B active per token, per reported figures — with a 1M-token context window, released under a plain MIT license. It was reportedly ranked the #1 open model on Artificial Analysis at launch, and unlike Kimi K3, it has a real local path today: the official Hugging Face model card confirms local-inference support, and 112 quantized versions are available for llama.cpp, Ollama, and LM Studio. The honest caveat: quantized, it still needs a big-RAM rig — Mac Studio 512GB class, an EPYC server with KTransformers, or a multi-GPU stack. This is not an 8GB-VRAM model.

📅 Published: July 20, 2026🔄 Last Updated: July 20, 2026✓ Manually Reviewed

Key takeaways

  • 753B total MoE, ~40B active (reported) — per-token compute is modest; memory is the constraint.
  • Plain MIT license — confirmed on the Hugging Face card. No custom riders.
  • Reported #1 open model at launch on Artificial Analysis.
  • 112 ready-made quants for llama.cpp / Ollama / LM Studio — the widest ready-made quant ecosystem of the June-July open wave.
  • Hardware floor: big-RAM rigs — ~380-420 GB at ~4-bit (estimate). Mac Studio 512GB, EPYC + KTransformers, or multi-GPU.
  • 1M-token context with Z.ai's new IndexShare sparse attention.

Quick verdict

GLM-5.2 is the model to pick if you want the biggest open model that can genuinely live on hardware you own. Most frontier-size open releases are "open" in the sense that a datacenter could run them. GLM-5.2 is different because the local tooling shipped with it: vLLM, SGLang, Transformers, KTransformers, and Unsloth support are confirmed on the Hugging Face card, and 112 quantized versions are already available covering llama.cpp, Ollama, and LM Studio.

Who it's not for: anyone on consumer GPUs. If your machine tops out at 24GB of VRAM or 64GB of RAM, look at smaller models — our Ollama RAM/VRAM table maps what fits where. And if you don't want to self-host at all, Z.ai's API runs $1.40/$4.40 per million input/output tokens (secondary-source pricing).

Specs at a glance

VendorZ.ai (Zhipu AI)
ReleasedHugging Face card dated June 17, 2026 (launch window June 13-16 per secondary sources)
ArchitectureMixture-of-Experts with IndexShare sparse attention
Total parameters753 billion
Active parameters~40 billion per token (reported by secondary sources)
Context window1,000,000 tokens
LicenseMIT (confirmed on HF card)
Local inferencevLLM, SGLang, Transformers, KTransformers, Unsloth; 112 quants for llama.cpp / Ollama / LM Studio
Hugging Facezai-org/GLM-5.2
API pricing$1.40 / $4.40 per Mtok in/out (secondary sources)

Can you actually run it locally?

Yes — with the caveat that "locally" here means a serious machine, not a gaming PC. What makes GLM-5.2 notable is that the local path is official and already turnkey. The Hugging Face card lists five supported inference stacks (vLLM, SGLang, Transformers, KTransformers, Unsloth), and 112 quantized versions already exist for llama.cpp, Ollama, and LM Studio. Compare that to Kimi K3 — 2.8T parameters, weights due around July 27 — which no home rig will ever hold.

The ~40B-active-per-token design (reported) is what makes this work. Once the weights fit in memory, each token only touches a fraction of them, so per-token compute stays modest. The whole problem collapses to one question: can your machine hold a few hundred gigabytes of weights with decent memory bandwidth? That's the question the next section answers.

Hardware requirements

Unquantized, 753B parameters is multi-GPU server territory — for scale, the predecessor GLM-5 needed roughly 8× H200 for FP8. Quantized, three realistic self-host targets emerge:

1. Mac Studio, 512GB unified memory

At ~4-bit the weights are roughly 380-420 GB (estimate) — inside a 512GB unified-memory envelope. The cleanest single-box path, using the ready-made llama.cpp / LM Studio quants.

2. EPYC / Threadripper server + KTransformers

384-512GB of system RAM with KTransformers (officially supported on the HF card). At 2-3-bit dynamic quants, weights shrink to roughly 200-300 GB (estimate), which is why the 384GB tier becomes viable.

3. Multi-GPU stack with large aggregate VRAM

vLLM or SGLang across enough cards to hold a quantized checkpoint. Highest bandwidth of the three options, and the highest cost.

What we won't do is quote you exact GGUF file sizes or tokens-per-second — those vary by quant and rig, and we haven't benchmarked this one ourselves yet. The sizes above are arithmetic estimates from the parameter count. If your budget is smaller than any of these tiers, the RAM/VRAM sizing table covers models that fit ordinary hardware.

GLM-5.2 vs the June-July open wave

GLM-5.2 landed in the middle of an unusually crowded open-weights stretch. How the wave stacks up:

ModelSizeReleasedLicenseLocal story
GLM-5.2753B / ~40B active (reported)June 2026MIT112 quants; llama.cpp / Ollama / LM Studio confirmed
MiniMax M3428B / ~23B activeJune 1Official MXFP8 + Unsloth GGUF
Nemotron 3 Ultra~550B / 55B activeJune 4OpenMDW
Tencent Hy3295B / 21B activeJuly 6Apache 2.0
Thinking Machines Inkling975BJuly 15
Kimi K32.8TWeights due ~July 27Not home-runnable

Two things distinguish GLM-5.2 in this crowd: it's the one reportedly ranked #1 open model on Artificial Analysis at launch, and it has the widest ready-made quant ecosystem — the only frontier-size release in the wave with confirmed Ollama and LM Studio coverage. Against DeepSeek V4, the other frontier open family, GLM-5.2's pitch is the same: plain MIT plus a turnkey local path. For the full landscape, see our open-source LLM rankings.

FAQ

What is GLM-5.2?
GLM-5.2 is Z.ai's (Zhipu AI's) flagship open-weight model, published on Hugging Face June 17, 2026 (secondary sources place the launch window June 13-16). It's a Mixture-of-Experts model with 753 billion total parameters — roughly 40B active per token, per reported figures — a 1-million-token context window, and a new sparse-attention scheme Z.ai calls IndexShare. It ships under a plain MIT license, confirmed on the Hugging Face model card, and was reportedly ranked the #1 open model on Artificial Analysis at launch.
Can I run GLM-5.2 locally with Ollama?
Yes — this is what separates GLM-5.2 from other frontier-size open releases. The Hugging Face card confirms local-inference support via vLLM, SGLang, Transformers, KTransformers, and Unsloth, with 112 quantized versions available for llama.cpp, Ollama, and LM Studio. The catch is hardware: at ~4-bit quantization the weights are roughly 380-420 GB (estimate), so you need a machine that can hold that in memory — a Mac Studio with 512GB unified memory, an EPYC/Threadripper server with 384-512GB RAM running KTransformers, or a multi-GPU stack with large aggregate VRAM. This is not an 8GB-VRAM model.
What hardware do I need for GLM-5.2?
Unquantized, GLM-5.2's 753B parameters put it in multi-GPU server territory — for scale, its predecessor GLM-5 needed roughly 8× H200 for FP8. Quantized, the realistic self-host targets are: a Mac Studio with 512GB unified memory, an EPYC or Threadripper server with 384-512GB of system RAM plus KTransformers, or a multi-GPU rig with large aggregate VRAM. Rough size arithmetic (estimates): ~380-420 GB at ~4-bit, ~200-300 GB at 2-3-bit dynamic quants. Because only ~40B parameters are active per token (reported), per-token compute is modest once the weights fit — memory capacity and bandwidth are the constraint, not FLOPs.
GLM-5.2 vs DeepSeek V4: which should I pick?
Both are frontier-class open-weight families. GLM-5.2's case: it was reportedly ranked #1 open model on Artificial Analysis at launch, it's plain MIT, and it has the widest ready-made quant ecosystem of the June-July open wave — 112 quantized versions covering llama.cpp, Ollama, and LM Studio, so the local path is turnkey rather than DIY. If you're choosing between them for self-hosting, start from what your hardware can hold and how much conversion work you're willing to do; see our DeepSeek V4 page for that model's specifics.
GLM-5.2 vs Kimi K3: what's the difference for local use?
Kimi K3 is far bigger — 2.8 trillion parameters, with weights due around July 27, 2026 — and it is not home-runnable in any practical sense. GLM-5.2 at 753B is the biggest open model with a genuine self-host path today: confirmed llama.cpp/Ollama/LM Studio support with 112 ready-made quants. If your goal is running a frontier-class open model on hardware you own, GLM-5.2 is the ceiling right now; K3 is an API-or-datacenter model.
Is GLM-5.2 free for commercial use?
Yes. The Hugging Face model card confirms a plain MIT license — no custom acceptable-use rider, no revenue thresholds, no restrictions on building competing services. That's simpler than the modified licenses attached to several other large open releases. If you'd rather not self-host, Z.ai's API is priced at $1.40 per million input tokens and $4.40 per million output tokens, per secondary sources.

Related models

  • GLM-5 — predecessor; the ~8× H200 FP8 baseline this page compares against
  • GLM-4.6 — earlier GLM generation
  • DeepSeek V4 — the other frontier open-weight family
  • MiniMax M3 — smaller wave-mate with official MXFP8 + Unsloth GGUF
  • Nemotron 3 — Nvidia's ~550B entry in the same wave
  • Kimi K3 — 2.8T; the model GLM-5.2 beats on self-hostability
  • Best open-source LLMs 2026 — pillar comparison
🎯
AI Learning Path

Go from reading about AI to building with AI

20 structured courses. Hands-on projects. Runs on your machine. Start free.

Or own it for life — Lifetime $149 $599, pay once
LM

Written by the Local AI Master Team

The team behind Local AI Master

We build Local AI Master around practical, testable local AI workflows: model selection, hardware planning, RAG systems, agents, and MLOps. The goal is to turn scattered tutorials into a structured learning path you can follow on your own hardware.

✓ Local AI Curriculum✓ Hands-On Projects✓ Open Source Contributor
More on Local AI vs Cloud
See the full Local AI vs Cloud AI guide.
📚
Free · no account required

Grab the AI Starter Kit — career roadmap, cheat sheet, setup guide

No spam. Unsubscribe with one click.

🎯
AI Learning Path

Found your model? Now build something with it.

20 hands-on courses — RAG, agents, fine-tuning — all running locally. First chapter free, no card.

Or own it for life — Lifetime $149 $599, pay once
Free Tools & Calculators