GLM · Over 80B · Mixture of experts
GLM 4.5 Air
110.47 billion parameters in total, but only 13.4 billion are read for each token. That split is the whole point of the design: it costs the memory of a large model and the speed of a small one. It uses grouped-query attention with 8 KV heads for 96 query heads, which cuts the cache to 8% of what full attention would need.
About GLM 4.5 Air
GLM 4.5 Air is a mixture-of-experts model with 110.47B parameters, 13.4B of them active for each token, built from 46 layers with a hidden size of 4096. Its repository was published on 2025-07-20. The license listed on Hugging Face is mit. It uses grouped-query attention with 8 KV heads for 96 query heads, which cuts the cache to 8% of what full attention would need.
Memory needed, by quantization
At an 8k context window. Quality is our rough ranking of how much the compression costs you: anything at or above Q5 is hard to tell apart from the original in normal use.
| Quantization | File size | KV cache | Total VRAM | Quality | What it costs you |
|---|---|---|---|---|---|
| Q8_0 | 109.31 GB | 1.44 GB | 110.54 GB | 99% | Lossless in practice. Use it when the memory is there. |
| Q6_K | 84.36 GB | 1.44 GB | 85.73 GB | 98% | Very close to Q8 for two thirds of the size. |
| Q5_K_M | 73.18 GB | 1.44 GB | 74.61 GB | 96% | The quality-first choice when Q6 will not fit. |
| Q4_K_M | 62.12 GB | 1.44 GB | 63.61 GB | 93% | The default. Best size-to-quality ratio for local use. |
| IQ4_XS | 54.66 GB | 1.44 GB | 56.19 GB | 91% | Importance-matrix 4-bit. Q4_K_S quality, smaller file. |
| Q3_K_M | 50.28 GB | 1.44 GB | 51.85 GB | 86% | Degradation starts to show. A way to fit one size up. |
| IQ3_XXS | 39.35 GB | 1.44 GB | 40.98 GB | 79% | Aggressive. Only worth it on very large models. |
What a longer conversation costs
Same model at Q4_K_M, only the context window changes. The cache grows in a straight line with every token in the window.
| Context | KV cache | Total VRAM | Fits in 8 GB | Fits in 12 GB | Fits in 24 GB |
|---|---|---|---|---|---|
| 4k | 0.72 GB | 62.89 GB | no | no | no |
| 8k | 1.44 GB | 63.61 GB | no | no | no |
| 16k | 2.88 GB | 65.05 GB | no | no | no |
| 32k | 5.75 GB | 67.92 GB | no | no | no |
| 64k | 11.5 GB | 73.67 GB | no | no | no |
| 128k | 23 GB | 85.17 GB | no | no | no |
The "fits" columns allow for the roughly 0.8 GB Windows keeps for the desktop.
Which hardware runs GLM 4.5 Air
7 of 118 consumer devices run it at a quantization worth using, at an 8k context window.
| Device | Memory | Verdict | Quantization | Used | Speed |
|---|---|---|---|---|---|
| Apple M3 Ultra 256GB Apple | 256 GB | Runs well | Q8_0 | 111.13 GB | 22.1 t/s |
| Apple M3 Ultra 512GB Apple | 512 GB | Runs well | Q8_0 | 111.13 GB | 22.1 t/s |
| Apple M2 Ultra 192GB Apple | 192 GB | Runs well | Q8_0 | 111.13 GB | 21.6 t/s |
| RTX PRO 6000 Blackwell NVIDIA | 96 GB | Runs great | Q6_K | 85.73 GB | 65.8 t/s |
| NVIDIA DGX Spark 128GB NVIDIA | 128 GB | Runs well | Q6_K | 86.21 GB | 10 t/s |
| Ryzen AI Max+ 395 96GB AMD | 96 GB | Runs well | IQ4_XS | 56.5 GB | 10.5 t/s |
| Ryzen AI Max+ 395 128GB AMD | 128 GB | Runs well | IQ4_XS | 56.5 GB | 10.5 t/s |
| Apple M4 Max 36GB Apple | 36 GB | Just fits | n/a ! | 28.31 GB | 46.5 t/s |
| Apple M1 Ultra 64GB Apple | 64 GB | Just fits | n/a | 56.47 GB | 39.3 t/s |
| Apple M2 Ultra 64GB Apple | 64 GB | Just fits | n/a | 56.47 GB | 39.3 t/s |
| Apple M4 Max 48GB Apple | 48 GB | Just fits | n/a ! | 41.17 GB | 34.8 t/s |
| Apple M1 Max 32GB Apple | 32 GB | Just fits | n/a ! | 28.31 GB | 34 t/s |
| Apple M2 Max 32GB Apple | 32 GB | Just fits | n/a ! | 28.31 GB | 34 t/s |
| Apple M3 Max 36GB Apple | 36 GB | Just fits | n/a ! | 28.31 GB | 34 t/s |
| Apple M3 Ultra 96GB Apple | 96 GB | Just fits | n/a | 86.18 GB | 27.8 t/s |
| Apple M4 Max 64GB Apple | 64 GB | Just fits | n/a | 56.47 GB | 26.8 t/s |
| Apple M3 Max 48GB Apple | 48 GB | Just fits | n/a ! | 41.17 GB | 25.5 t/s |
| Apple M1 Ultra 128GB Apple | 128 GB | Just fits | n/a | 111.13 GB | 21.6 t/s |
| Apple M2 Ultra 128GB Apple | 128 GB | Just fits | n/a | 111.13 GB | 21.6 t/s |
| Apple M1 Max 64GB Apple | 64 GB | Just fits | n/a | 56.47 GB | 19.6 t/s |
| Apple M2 Max 64GB Apple | 64 GB | Just fits | n/a | 56.47 GB | 19.6 t/s |
| Apple M3 Max 64GB Apple | 64 GB | Just fits | n/a | 56.47 GB | 19.6 t/s |
| Jetson AGX Orin 32GB NVIDIA | 32 GB | Just fits | n/a ! | 28.34 GB | 18.9 t/s |
| Ryzen AI Max+ 395 32GB AMD | 32 GB | Just fits | n/a ! | 28.34 GB | 18.2 t/s |
| Apple M4 Pro 48GB Apple | 48 GB | Just fits | n/a ! | 41.17 GB | 17.4 t/s |
Show 35 more devices
| Apple M1 Pro 32GB Apple | 32 GB | Just fits | n/a ! | 28.31 GB | 17 t/s |
| Apple M2 Pro 32GB Apple | 32 GB | Just fits | n/a ! | 28.31 GB | 17 t/s |
| Apple M4 Max 128GB Apple | 128 GB | Just fits | n/a | 111.13 GB | 14.7 t/s |
| Apple M2 Max 96GB Apple | 96 GB | Just fits | n/a | 86.18 GB | 13.6 t/s |
| Apple M3 Max 96GB Apple | 96 GB | Just fits | n/a | 86.18 GB | 13.6 t/s |
| Apple M4 Pro 64GB Apple | 64 GB | Just fits | n/a | 56.47 GB | 13.4 t/s |
| Apple M5 32GB Apple | 32 GB | Just fits | n/a ! | 28.31 GB | 13 t/s |
| Apple M3 Pro 36GB Apple | 36 GB | Just fits | n/a ! | 28.31 GB | 12.8 t/s |
| Jetson AGX Orin 64GB NVIDIA | 64 GB | Just fits | n/a | 56.5 GB | 10.9 t/s |
| Apple M3 Max 128GB Apple | 128 GB | Just fits | n/a | 111.13 GB | 10.8 t/s |
| Ryzen AI Max+ 395 64GB AMD | 64 GB | Just fits | n/a | 56.5 GB | 10.5 t/s |
| Apple M4 32GB Apple | 32 GB | Just fits | n/a ! | 28.31 GB | 10.2 t/s |
| Intel Core Ultra 9 288V 32GB Intel | 32 GB | Just fits | n/a ! | 28.34 GB | 9.7 t/s |
| Ryzen AI 9 HX 370 32GB AMD | 32 GB | Just fits | n/a ! | 28.34 GB | 9.1 t/s |
| RTX 6000 Ada Generation NVIDIA | 48 GB | Partial offload 46/46 | n/a | 63.61 GB | 18.9 t/s |
| RTX A6000 NVIDIA | 48 GB | Partial offload 46/46 | n/a | 63.61 GB | 17.4 t/s |
| Radeon PRO W7900 AMD | 48 GB | Partial offload 46/46 | n/a | 63.61 GB | 17.1 t/s |
| GeForce RTX 5090 NVIDIA | 32 GB | Partial offload 46/46 | n/a | 63.61 GB | 13.1 t/s |
| Radeon RX 7900 XTX AMD | 24 GB | Partial offload 46/46 | n/a | 56.19 GB | 11.7 t/s |
| RTX A5000 NVIDIA | 24 GB | Partial offload 46/46 | n/a | 56.19 GB | 11.6 t/s |
| GeForce RTX 5080 Laptop NVIDIA | 16 GB | Partial offload 46/46 | n/a | 51.85 GB | 10.9 t/s |
| GeForce RTX 4080 SUPER NVIDIA | 16 GB | Partial offload 46/46 | n/a | 51.85 GB | 10.8 t/s |
| GeForce RTX 4080 NVIDIA | 16 GB | Partial offload 46/46 | n/a | 51.85 GB | 10.8 t/s |
| GeForce RTX 4070 Ti SUPER NVIDIA | 16 GB | Partial offload 46/46 | n/a | 51.85 GB | 10.7 t/s |
| Radeon RX 7900 XT AMD | 20 GB | Partial offload 46/46 | n/a | 56.19 GB | 10.4 t/s |
| GeForce RTX 3080 Ti NVIDIA | 12 GB | Partial offload 46/46 | n/a | 51.85 GB | 10.3 t/s |
| GeForce RTX 3080 12GB NVIDIA | 12 GB | Partial offload 46/46 | n/a | 51.85 GB | 10.3 t/s |
| GeForce RTX 4090 Laptop NVIDIA | 16 GB | Partial offload 46/46 | n/a | 51.85 GB | 10.3 t/s |
| GeForce RTX 5080 NVIDIA | 16 GB | Partial offload 46/46 | n/a | 56.19 GB | 10.2 t/s |
| Radeon RX 9070 XT AMD | 16 GB | Partial offload 46/46 | n/a | 51.85 GB | 10.2 t/s |
| Radeon RX 9070 AMD | 16 GB | Partial offload 46/46 | n/a | 51.85 GB | 10.2 t/s |
| Radeon RX 7800 XT AMD | 16 GB | Partial offload 46/46 | n/a | 51.85 GB | 10.2 t/s |
| GeForce RTX 5070 Ti NVIDIA | 16 GB | Partial offload 46/46 | n/a | 56.19 GB | 10.1 t/s |
| GeForce RTX 4090 NVIDIA | 24 GB | Partial offload 46/46 | n/a | 63.61 GB | 10.1 t/s |
| GeForce RTX 3090 Ti NVIDIA | 24 GB | Partial offload 46/46 | n/a | 63.61 GB | 10.1 t/s |
How it compares
Other models in the same family or of a similar size, on a 12 GB card and a 24 GB card, each at the best quantization that fits.
| Model | Weights at 4 bits | GeForce RTX 3060 12GB | GeForce RTX 4090 |
|---|---|---|---|
| GLM 4.5 Air | 62.1 GB | does not fit well | 10.1 t/s at Q4_K_M |
| GLM 4.7 Flash | 17.6 GB | 37.6 t/s at Q4_K_M | 152.9 t/s at Q5_K_M |
| GLM 4.5 | 201.5 GB | does not fit well | does not fit well |
| GLM 5.2 | 423.6 GB | does not fit well | does not fit well |
| GPT OSS 120B | 59.0 GB | 16.9 t/s at MXFP4 | 23.9 t/s at MXFP4 |
| Qwen3.5 122B A10B | 70.3 GB | 12.6 t/s at Q4_K_M | 17.7 t/s at Q4_K_M |
What to buy to run GLM 4.5 Air
The cheapest hardware that runs it at a quantization worth using and a speed you would not resent, at an 8k context window.
Cheapest that works
Apple M1 Max 64GB
$1,900 whole machine
IQ4_XS · uses 56.47 GB of its 64 GB · about 19.6 tokens/s
Best value
Apple M1 Ultra 64GB
$2,400 whole machine
IQ4_XS · about 39.3 tokens/s · 800 GB/s
If it just has to run
GeForce GTX 1660 SUPER
$110 used
IQ4_XS · about 7.2 tokens/s. Cheaper, and slower to answer.
| Device | Price | Memory | Runs it at | Speed | |
|---|---|---|---|---|---|
| Apple M1 Max 64GB Apple · whole machine | $1,900 | 64 GB | IQ4_XS | 19.6 t/s | Check price |
| Apple M1 Ultra 64GB Apple · whole machine | $2,400 | 64 GB | IQ4_XS | 39.3 t/s | Check price |
| Apple M1 Ultra 128GB Apple · whole machine | $3,400 | 128 GB | Q8_0 | 21.6 t/s | Check price |
| Apple M3 Ultra 256GB Apple · whole machine | $5,599 | 256 GB | Q8_0 | 22.1 t/s | Check price |
| RTX PRO 6000 Blackwell NVIDIA | $8,500 | 96 GB | Q6_K | 65.8 t/s | Check price |
| Apple M3 Ultra 512GB Apple · whole machine | $9,499 | 512 GB | Q8_0 | 22.1 t/s | Check price |
Indicative prices reviewed 2026-09-01; used prices are marketplace typical. Options that cost more than a cheaper one with no more memory and no more speed are hidden. Price links are Amazon affiliate links. Change the standard, the context window or the budget in the buying tool.
Where it sits in the catalog
Source: zai-org/GLM-4.5-Air on Hugging Face. Downloaded 116k times in the last month. Published 2025-07-20. Architecture figures are read from the repository's own configuration file, so they move when the model does.
Questions people ask about GLM 4.5 Air
How much VRAM does GLM 4.5 Air need?
At Q4_K_M with an 8k context window, GLM 4.5 Air needs about 63.61 GB: 61.77 GB of weights, 1.44 GB of KV cache and 0.41 GB of overhead. At the highest-quality file it needs about 110.54 GB.
What is the cheapest GPU that runs GLM 4.5 Air?
The smallest consumer device in our catalog that fits it is the Ryzen AI Max+ 395 96GB with 96 GB, running it at IQ4_XS at roughly 10.5 tokens per second.
How much does a longer context window add for GLM 4.5 Air?
On the memory side, a 32k window adds 5.8 GB of KV cache on top of the weights, and the cache grows in a straight line with every token. The longest window the model supports natively is 128k tokens.