GLM-4.5-Air
GLM-4.5-Air by Zhipu AI (Z.ai) (106B parameters, MoE with 12B active per token). At Q4_K_M with an 8K context it needs about 64.8 GiB of memory. Z.ai's 106B MoE (12B active) with hybrid thinking, under MIT; at 4-bit it needs a 72 GB card or a 96 GB unified-memory machine. There is no official Ollama tag.
Last updated: 2026-10-13
Model facts
Memory needed
Weights + KV cache + runtime overhead, in GiB. Lower quantization = less memory, slightly lower quality.
| Quantization | 4K context | 8K context | 32K context |
|---|---|---|---|
| Q4_K_M | 64.1 GiB | 64.8 GiB | 69.1 GiB |
| Q5_K_M | 74.9 GiB | 75.7 GiB | 80.0 GiB |
| Q6_K | 86.6 GiB | 87.3 GiB | 91.6 GiB |
| Q8_0 | 111.4 GiB | 112.1 GiB | 116.4 GiB |
| FP16 | 208.5 GiB | 209.2 GiB | 213.6 GiB |
Which hardware runs it
Best quantization that fits each device with a 8K context. Estimates, not benchmarks. See the methodology.
Frequently asked questions
How much VRAM does GLM-4.5-Air need?
With an 8K context, about 64.8 GiB at Q4_K_M and 112.1 GiB at Q8_0 (weights + KV cache + runtime overhead). Longer contexts need more memory for the KV cache.
Can GLM-4.5-Air run on an GeForce RTX 4090?
GLM-4.5-Air does not fit on the GeForce RTX 4090 (24 GB), even at Q4_K_M with an 8K context.
Can I run GLM-4.5-Air on a Mac?
Yes. GLM-4.5-Air fits on the M3 Ultra (96 GB) at Q4_K_M with an 8K context, using about 64.8 GiB, at an estimated 73 tokens/s.
Can GLM-4.5-Air be used commercially?
GLM-4.5-Air is released under MIT. The licence allows commercial use (read its conditions).