back to local model checker

GLM-4.6

MoETools

Z.ai · 357B (32B active) · Mixture of Experts

Large GLM MoE with strong coding and 200K context

MIT
Downloads

0

Likes

0

Released

2025-09

Context

195K tokens

Use cases

chatcodereasoning
Total experts160
Active experts8
Active params32.0B

Your hardware

16 GB Reference Desktop

Quant
Q4_K_M
Memory
183.4 GB
Speed
--
Fit
Too heavy

Quantization options

What is quantization?
QuantBitsVRAMDiskQualityStatus
Q2_K2114.8 GB109.1 GBlowToo heavy · ~1 t/s
Q3_K_M3160.5 GB152.7 GBmoderateToo heavy · ~1 t/s
Q4_K_M4183.4 GB174.6 GBgoodToo heavy · ~1 t/s
Q5_K_M5229.1 GB218.2 GBgoodToo heavy · ~1 t/s
Q6_K6274.8 GB261.8 GBexcellentToo heavy · ~1 t/s
Q8_08366.2 GB349.1 GBexcellentToo heavy · ~1 t/s
F1616732 GB698.2 GBlosslessToo heavy · ~1 t/s

About this model

GLM-4.6 (GLM-4.6) is a 357B sparse mixture-of-experts (MoE) model activating 32B active per token from Z.ai in the GLM family, designed primarily for chat, code, reasoning workflows. Supports structured tool calling; use Q4_K_M or higher when strict JSON schema adherence matters. Supports up to 195K context tokens; start at 4K–8K with Q8_0 KV cache on tight GPUs.

Hardware planning baselines start at 199.5 GB minimum memory and 332.5 GB recommended memory. At Q4_K_M, weights plus baseline runtime overhead use roughly 183.4 GB of RAM/VRAM and 174.6 GB of disk space. At 357B, GLM-4.6 belongs firmly in the workstation and multi-GPU tier. Running it comfortably requires 332.5 GB of total memory across dual RTX GPUs, high-memory Apple Silicon (Mac Studio Ultra/Max), or MoE hybrid CPU-RAM offloading—mobile devices and standard 16 GB laptops are too constrained for practical speeds.

Deploy GLM-4.6 locally via Ollama (`glm-4.6`), Ollama with Flash Attention enabled. Under MIT License, review official weight cards before commercial redistribution.

Highlights

  • Local Fit Baseline199.5 GB minimum and 332.5 GB recommended RAM/VRAM.
  • Primary QuantQ4_K_M targets 183.4 GB memory and 174.6 GB disk.
  • Context & Capabilities195K max context · Sparse MoE · Tool calling.

System requirements by device

Memory planning targets 199.5 GB minimum RAM/VRAM and 332.5 GB recommended memory. The baseline quant is Q4_K_M, estimated at 183.4 GB memory and 174.6 GB disk. Actual runtime usage shifts with context length, KV cache precision, and OS background overhead.

DeviceNotes
Mac (Apple silicon)Requires a 48 GB–192 GB Mac Studio or MacBook Pro Max/Ultra with enough unified wired budget for 183.4 GB.
Windows / Linux GPUUse dual/multi-GPU CUDA splitting (--tensor-split) or MoE CPU expert offloading (-ot) paired with 32–128 GB DDR5 system RAM.
Phones & Compact Edge BoardsNot practical on phones (199.5 GB min RAM exceeds Android LMK and iOS Jetsam limits). Pick a 2B–4B model on mobile.

Start with a short session and leave memory for the operating system. A file fitting on disk does not guarantee that it will load in memory.

Estimated memory by quantization

These are rough estimates, not tested minimum requirements. Check the actual file and runtime documentation; context, offloading and operating-system overhead can change the memory needed.

Quant TierEstimated memoryDisk SizeQuality
Q2_K114.8 GB109.1 GBlow
Q3_K_M160.5 GB152.7 GBmoderate
Q4_K_M183.4 GB174.6 GBgood
Q5_K_M229.1 GB218.2 GBgood
Q6_K274.8 GB261.8 GBexcellent
Q8_0366.2 GB349.1 GBexcellent
F16732.0 GB698.2 GBlossless

How to run GLM-4.6 on your device

Start with a short prompt and the smallest quant that still gives the quality you need. A clean first run is more useful than loading a larger file with no memory headroom.

Windows

Run GLM-4.6 on Multi-GPU or High-RAM Desktop

  1. Plan for 199.5 GB minimum combined memory and 332.5 GB recommended memory for Q4_K_M.
  2. Use llama.cpp `-ot ".ffn_.*_exps.=CPU"` or `--tensor-split` across multiple GPUs so attention and KV cache remain in fast VRAM.
  3. Enable Flash Attention (`-fa on`) and Q8_0 KV cache (`-ctk q8_0 -ctv q8_0`) to prevent context buffers from exceeding GPU headroom.
  4. Avoid attempting to run workstation-scale weights on smartphones or 8 GB laptops; use a 2B–8B checkpoint on portable devices.
Device how-to guide
Mac

Run GLM-4.6 on High-Memory Apple Silicon Mac

  1. Verify you are on a 48 GB–192 GB Mac Studio or MacBook Pro Max/Ultra capable of holding 183.4 GB in wired unified memory.
  2. Launch via `llama-server` or LM Studio with MLX/Metal enabled and set initial context conservatively to 4K–8K.
  3. Monitor Activity Monitor Memory Pressure to confirm zero SSD swap thrashing during token generation.
  4. Step down to a lighter quantization (e.g., Q3_K_M or IQ4_XS) if total memory utilization exceeds 75%.
Device how-to guide

GLM-4.6 vs alternatives

ModelParamsContextMin RAMLicense
GLM-4.6357B (32B active)195K199.5 GBMIT
GLM-4.5 Air106B (12B active)128K59.2 GBMIT
GLM-4 9B9B128K5 GBGLM-4
GLM-5744B (40B active)128K415.7 GBMIT
GLM-5.2753B (40B active)1M420.8 GBMIT
GLM-5.3753B (40B active)1M420.8 GBMIT

Use this table to choose the smallest nearby model that still matches the job. A related model with lower RAM can be the better local choice when the current page fits only at the edge of available memory.

Frequently asked questions

What is GLM-4.6?
GLM-4.6 is a 357B sparse mixture-of-experts model (32B active active per token) from Z.ai, supporting up to 195K tokens with 199.5 GB minimum RAM and 332.5 GB recommended RAM.
How do I download and install GLM-4.6?
Download a GGUF build for GLM-4.6, start with Q4_K_M, and confirm the model fits locally before increasing context or quant quality.
How much VRAM or RAM does GLM-4.6 require?
At Q4_K_M, GLM-4.6 occupies roughly 183.4 GB of active memory and 174.6 GB on disk. Add 2–6 GB of spare headroom for KV cache buffers, runtime scratch space, and OS apps (332.5 GB recommended).
What is the best way to run GLM-4.6 locally?
Use `ollama run glm-4.6` for a quick CLI/API setup, or load Q4_K_M in LM Studio or llama.cpp with Flash Attention enabled when you want explicit KV cache and layer control.
Can GLM-4.6 run on a consumer laptop or smartphone?
No smartphone or 8–16 GB laptop can hold GLM-4.6 cleanly in local memory. It targets multi-GPU NVIDIA workstations, high-RAM Mac Studios, or MoE CPU-RAM hybrid rigs with at least 199.5 GB total memory.
How does GLM-4.6 compare to GLM-4.5 Air?
GLM-4.5 Air (106B, 59.2 GB min RAM) sits closest in size. Compare active parameters, context ceiling (195K vs 128K), and license terms (MIT) before switching.