Run GLM-4.5-Air on-premises
○ Memory-fit arithmetic — checkable math, not a benchmark
GLM-4.5-Air is the lighter member of the GLM-4.5 family — 106 billion total parameters with about 12 billion active — released under the permissive MIT license and tuned for agentic and coding work. It fits a 128 GB machine at 4-bit.
Here is the arithmetic against a machine with 128 GB of memory.
Memory fit in 128 GB
| Parameters | 106B total, ~12B active per token (MoE) |
|---|---|
| 4-bit weights | ~55 GB |
| Reference memory | 128 GB (one machine) |
| Left for context cache + system | roughly 55 GB |
| Fits in 128 GB | yes |
At ~55 GB the model fits a 128 GB machine with generous context headroom; the full GLM-4.5 (~355B) does not — see the limits.
What it serves well
- Agentic and tool-calling workflows on your own machine.
- Coding assistance where an MIT license is a clean fit.
- General assistant duty with a permissive, commercial-friendly license.
Honest limits
- Memory arithmetic, not a benchmark. The ~55 GB envelope leaves comfortable room in 128 GB.
- The full GLM-4.5 / 4.6 (~355B) does NOT fit in 128 GB of memory — this page is specifically about the Air variant.
Frequently asked questions
Does GLM-4.5-Air fit in 128 GB of memory?
Yes — ~55 GB at 4-bit against 128 GB leaves roughly 55 GB for context and system.
License?
MIT — permissive and commercial-friendly, verified against the model's release.
Can I run the full GLM-4.5 in 128 GB?
No — the full 355B model exceeds 128 GB of memory and needs a GPU server or a private cloud. The Air variant on this page is the one that fits.
Find out which model fits your work
An AI diagnostic maps your processes, classes what is sensitive and sizes the model and the hardware against your real tasks.