Skip to content

GLM 4.5 Air VRAM requirements

How much VRAM GLM 4.5 Air needs
4-bit weights: 62.4 GBWith 8K context: 64.7 GBWith 32K context: 69.0 GB8-bit, 8K context: 112 GB
Parameters
110B
Active per token
15B
Longest context
128K
Released
Jul 2025
License
MIT
Strongest at
Not rated

How much VRAM GLM 4.5 Air needs

Memory GLM 4.5 Air needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
16-bit206 GB208 GB212 GB230 GB
8-bit109 GB112 GB116 GB133 GB
6-bit84.4 GB86.7 GB91.0 GB108 GB
5-bit73.2 GB75.5 GB79.8 GB97.0 GB
4-bit62.4 GB64.7 GB69.0 GB86.2 GB
3-bit50.3 GB52.6 GB56.9 GB74.1 GB
2-bit38.6 GB40.9 GB45.2 GB62.4 GB
1.6-bit21.2 GB23.5 GB27.8 GB45.1 GB

One conversation, 16-bit context cache, runtime overhead included.

GPUs and Macs that run GLM 4.5 Air

The best way GLM 4.5 Air runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: GLM 4.5 Air on RTX 4060 report
RTX 5060Too bigPDF: GLM 4.5 Air on RTX 5060 report
RTX 3070Too bigPDF: GLM 4.5 Air on RTX 3070 report
RTX 3080 10 GBToo bigPDF: GLM 4.5 Air on RTX 3080 10 GB report
RTX 3060 12 GBToo bigPDF: GLM 4.5 Air on RTX 3060 12 GB report
Arc B580Too bigPDF: GLM 4.5 Air on Arc B580 report
RTX 4070Too bigPDF: GLM 4.5 Air on RTX 4070 report
RTX 4070 SuperToo bigPDF: GLM 4.5 Air on RTX 4070 Super report
RTX 5070Too bigPDF: GLM 4.5 Air on RTX 5070 report
RTX 4060 Ti 16 GBToo bigPDF: GLM 4.5 Air on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GBToo bigPDF: GLM 4.5 Air on RTX 5060 Ti 16 GB report
Arc A770 16 GBToo bigPDF: GLM 4.5 Air on Arc A770 16 GB report
Radeon RX 9070 XTToo bigPDF: GLM 4.5 Air on Radeon RX 9070 XT report
RTX 4070 Ti SuperToo bigPDF: GLM 4.5 Air on RTX 4070 Ti Super report
RTX 4080 SuperToo bigPDF: GLM 4.5 Air on RTX 4080 Super report
RTX 5070 TiToo bigPDF: GLM 4.5 Air on RTX 5070 Ti report
RTX 5080Too bigPDF: GLM 4.5 Air on RTX 5080 report
Radeon RX 7900 XTXToo bigPDF: GLM 4.5 Air on Radeon RX 7900 XTX report
RTX 3090Too bigPDF: GLM 4.5 Air on RTX 3090 report
RTX 4090Too bigPDF: GLM 4.5 Air on RTX 4090 report
Radeon AI PRO R9700Too bigPDF: GLM 4.5 Air on Radeon AI PRO R9700 report
RTX 5090Too bigPDF: GLM 4.5 Air on RTX 5090 report
2× RTX 30902-bit28K88 tok/s1.6 sPDF: GLM 4.5 Air on 2× RTX 3090 report
2× RTX 40902-bit28K93 tok/s0.7 sPDF: GLM 4.5 Air on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB2-bit32K115 tok/s1.5 sPDF: GLM 4.5 Air on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada2-bit32K90 tok/s1.1 sPDF: GLM 4.5 Air on RTX 6000 Ada report
L40S2-bit32K83 tok/s1.8 sPDF: GLM 4.5 Air on L40S report
2× RTX 50903-bit48K116 tok/s0.6 sPDF: GLM 4.5 Air on 2× RTX 5090 report
4× RTX 30906-bit20K47 tok/s0.9 sPDF: GLM 4.5 Air on 4× RTX 3090 report
RTX PRO 6000 Blackwell6-bit32K80 tok/s0.8 sPDF: GLM 4.5 Air on RTX PRO 6000 Blackwell report
Mac mini M4 16 GBToo bigPDF: GLM 4.5 Air on Mac mini M4 16 GB report
Mac mini M4 24 GBToo bigPDF: GLM 4.5 Air on Mac mini M4 24 GB report
Mac mini M4 32 GBToo bigPDF: GLM 4.5 Air on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GBToo bigPDF: GLM 4.5 Air on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GBToo bigPDF: GLM 4.5 Air on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB2-bit32K32 tok/s32 sPDF: GLM 4.5 Air on Mac mini M4 Pro 64 GB report
M4 Max 128 GB6-bit32K27 tok/s16 sPDF: GLM 4.5 Air on M4 Max 128 GB report
M5 Max 128 GB6-bit32K30 tok/s4.7 sPDF: GLM 4.5 Air on M5 Max 128 GB report
M3 Ultra 96 GB4-bit28K43 tok/s9.0 sPDF: GLM 4.5 Air on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit128K28 tok/s9.0 sPDF: GLM 4.5 Air on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit128K28 tok/s9.0 sPDF: GLM 4.5 Air on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit128K37 tok/s2.8 sPDF: GLM 4.5 Air on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB2-bit32K25 tok/s9.3 sPDF: GLM 4.5 Air on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB6-bit32K12 tok/s9.3 sPDF: GLM 4.5 Air on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit20K13 tok/s3.2 sPDF: GLM 4.5 Air on RTX Spark report
DGX Spark8-bit20K12 tok/s3.2 sPDF: GLM 4.5 Air on DGX Spark report
Dual DGX Spark8-bit128K12 tok/s1.9 sPDF: GLM 4.5 Air on Dual DGX Spark report
DGX Station GB3008-bit128K168 tok/s0.3 sPDF: GLM 4.5 Air on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit128K65 tok/s0.4 sPDF: GLM 4.5 Air on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit128K65 tok/s0.2 sPDF: GLM 4.5 Air on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit128K65 tok/s0.1 sPDF: GLM 4.5 Air on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM4-bit64K110 tok/s2.1 sPDF: GLM 4.5 Air on A100 80 GB SXM report
H100 SXM4-bit64K151 tok/s0.6 sPDF: GLM 4.5 Air on H100 SXM report
4× H200 NVL8-bit128K133 tok/s0.2 sPDF: GLM 4.5 Air on 4× H200 NVL report
8× H200 NVL8-bit128K133 tok/s0.1 sPDF: GLM 4.5 Air on 8× H200 NVL report

Best fit is the highest quality, up to 8-bit, that leaves 16K of context. First reply is the wait for a 2K prompt. Speeds are rough, plus or minus 25%.

How to run it

llama.cpp
llama-server -hf unsloth/GLM-4.5-Air-GGUF
vLLM
vllm serve zai-org/GLM-4.5-Air
MLX (Mac)
mlx_lm.chat --model mlx-community/GLM-4.5-Air-4bit

Questions

How much VRAM does GLM 4.5 Air need?

GLM 4.5 Air needs 62.4 GB for its weights at 4-bit, 64.7 GB with an 8K context and 69.0 GB with 32K. At 8-bit, 112 GB.

What is the minimum VRAM for GLM 4.5 Air?

About 40.9 GB at 2-bit with an 8K context, the lowest precision we recommend for a model this size. The smallest machine here that runs it without spilling is the 2× RTX 3090.

How big is GLM 4.5 Air?

110B parameters (15B active per token): 206 GB of weights at 16-bit, 62.4 GB at 4-bit.

Can GLM 4.5 Air run on an RTX 4090?

Not on one card. The smallest machine here that runs it is the 2× RTX 3090.

More models