Skip to content

Text model · GLM

GL GLM-4.5-Air: RAM and VRAM requirements

GLM-4.5-Air needs about 71.8 GB to run at Q4_K_M (the Q4_K_M GGUF file is ~68.45 GB to download; KV cache and overhead add the rest), or about 112.7 GB at Q8_0. The lightest hardware that runs it is Apple M4 Max (128GB).

GLM family · 106B params (Mixture-of-Experts: activates only 12B of 106B params per token, so generation is faster than the total size suggests) · released Jul 2025.

Shopping for hardware? See what runs GLM-4.5-Air →

License MIT · Commercial OK ↓ 441.3K/mo ♥ 625 on HuggingFace
Q4_K_M GGUF
68.45 GB
Q8_0 GGUF
109.39 GB
Memory @ Q4 (4k)
~71.8 GB
Context
128 k

Q4_K_M and Q8_0 are GGUF download (file) sizes. Memory @ Q4 includes the KV cache and overhead, so it runs larger than the file on disk.

Will it run on your device?

GLM-4.5-Air runs on 4 of 40 tracked devices at Q4_K_M.

4 run well 0 tight fit 36 too small
Fit check Q4_K_M
No, not enough memory
needs 71.8 GB usable 10.5 GB

Memory breakdown

Weights (Q4_K_M)68.45 GB
+
KV cache (4k)2.5 GB
+
Overhead0.8 GB
=
Total71.8 GB

How context length changes it

4k context ~71.8 GB 32k context ~89.1 GB 128k context ~148.4 GB

Longer context grows the KV cache, which for GLM-4.5-Air is sized by its 12B active params, not the full 106B. It needs ~71.8 GB at 4k and ~148.4 GB at 128k.

Quantization sizes

GGUF quantson disk
Quantization Size on disk
Q2_K 44.4 GB est
Q3_K_M 51.8 GB est
Q4_K_M (default) 68.45 GB
Q5_K_M 75.5 GB est
Q6_K 86.9 GB est
Q8_0 109.39 GB
FP16 212 GB est

Lower quant = smaller and faster, slightly lower quality. Q4_K_M is the common default.

Run it

llama.cpp
$ llama-cli -hf bartowski/zai-org_GLM-4.5-Air-GGUF:Q4_K_M
LM Studio
$ lms get bartowski/zai-org_GLM-4.5-Air-GGUF

Which devices can run GLM-4.5-Air?

Same job, different size

Pick by what fits your memory: step down to free up VRAM, or step up if you have headroom.

Similar models

Head-to-head

FAQ

How much VRAM or RAM does GLM-4.5-Air need?

At Q4_K_M, GLM-4.5-Air needs about 71.8 GB (weights ~68.45 GB + KV cache + overhead) at a 4k context. At Q8_0 budget ~112.7 GB.

What is the Q4_K_M GGUF file size of GLM-4.5-Air?

The Q4_K_M GGUF file is about 68.45 GB to download, and the Q8_0 GGUF is about 109.39 GB. That is the weights file on disk; to run it you also need room for the KV cache and overhead, so budget ~71.8 GB of memory at Q4_K_M.

Can GLM-4.5-Air run on a laptop?

GLM-4.5-Air is large; you need a high-memory Mac or multi-GPU setup at Q4_K_M.

Is GLM-4.5-Air cheaper to run because it is a MoE model?

It is faster, not lighter. GLM-4.5-Air activates only 12B of 106B params per token (so it runs quickly), but all experts must stay in memory, so it still needs memory for the full 106B.

Can I use GLM-4.5-Air commercially?

Yes. GLM-4.5-Air is licensed MIT, which permits commercial use.

Understand the numbers

Short guides to the ideas behind GLM-4.5-Air's memory and quant figures.

MoE: 128 routed experts + 1 shared expert, activates 9 per token. 106B total / 12B active per zai-org HF card and artificialanalysis.ai. Context 128K (max_position_embeddings=131072) per search results and llm-stats. MIT license, commercial use explicitly permitted per zai-org HF card. Released July 28 2025 per artificialanalysis.ai and llm-stats. Q4_K_M 73.50 GB and Q8_0 117.46 GB measured from bartowski GGUF file listing. fp16_gb not listed in bartowski repo (only goes to Q8_0 at 117.46 GB). No official ollama.com/library tag; community tag MichelRosselli/GLM-4.5-Air exists but is unofficial.

Sources

Last validated 2026-08-03. Memory figures are estimates. See methodology.