Can I run GLM 4.5 Air locally?
GLM 4.5 Air by zai-org has 110B parameters (15B active per token, 128 experts, 8 used per token). At 4-bit with an 8K context it needs about 70.1 GB of GPU memory. It fits on 22 of 65 GPUs and Macs we track.
Memory needed by quant and context length
GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from unsloth/GLM-4.5-Air-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.
| Quant | Quality | 4K ctx | 8K ctx | 32K ctx | 128K ctx |
|---|---|---|---|---|---|
| 2-bit (Q2_K) • | rough | 43.6 GB | 44.3 GB | 48.6 GB | 65.9 GB |
| 3-bit (Q3_K_M) • | usable | 54.7 GB | 55.4 GB | 59.7 GB | 77.0 GB |
| 4-bit (Q4_K_M) • | good, most popular | 69.4 GB | 70.1 GB | 74.4 GB | 91.7 GB |
| 5-bit (Q5_K_M) • | very good | 79.2 GB | 79.9 GB | 84.2 GB | 101 GB |
| 6-bit (Q6_K) • | near lossless | 93.6 GB | 94.3 GB | 98.6 GB | 116 GB |
| 8-bit (Q8_0) • | effectively lossless | 111 GB | 112 GB | 116 GB | 133 GB |
| 16-bit (F16/BF16) • | original weights | 207 GB | 208 GB | 212 GB | 230 GB |
Which GPUs and Macs run GLM 4.5 Air?
Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).
Prefer an API?
Z.ai: GLM 4.5 Air is on OpenRouter at $0.13 per 1M input tokens and $0.85 per 1M output tokens. Price details and history →
Questions
How much VRAM does GLM 4.5 Air need?
About 70.1 GB at 4-bit (Q4_K_M) with an 8K context: 68.0 GB of weights (real GGUF file size), 1.4 GB of KV cache and 0.7 GB of runtime overhead. At 8-bit it needs about 112 GB.
What is the smallest GPU that runs GLM 4.5 Air?
No single consumer GPU in our table fits it at Q4_K_M with 8K context; the smallest option that does is the RTX PRO 6000 Blackwell (96 GB).
Can I run GLM 4.5 Air on a Mac?
Yes. At Q4_K_M with 8K context it fits on a Mac M3 Max 128GB or larger, using the default macOS GPU memory limit.
Similar models
Same family or a similar parameter count, with the VRAM each needs at Q4_K_M and 8K context.
GLM 4.5V
108B · 61.4 GB at 4-bitGLM 4.6V
108B · 67.8 GB at 4-bitGLM Z1 32B 0414
33B · 19.5 GB at 4-bitLlama 4 Scout 17B 16E Instruct
109B · 63.2 GB at 4-bitLlama 4 Scout 17B 16E
109B · 64.1 GB at 4-bitgpt oss 120b
117B · 59.5 GB at 4-bitModel facts
| Hugging Face repo | zai-org/GLM-4.5-Air |
| Parameters | 110,468,824,832 (110B) · ~15B active (computed from config) |
| Layers / hidden size | 46 / 4096 |
| Attention | GQA; 96 heads, 8 KV heads, head dim 128 |
| KV cache per token (FP16) | 184.0 KiB |
| Max context | 131,072 tokens |
| License | mit |
| Released on HF | Jul 20, 2025 |
| Downloads (30d) / likes | 1,115,331 / 650 |
Sources: Hugging Face model API · config.json (zai-org/GLM-4.5-Air) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.