Skip to content

GLM 4.5 VRAM requirements

How much VRAM GLM 4.5 needs
4-bit weights: 202 GBWith 8K context: 206 GBWith 32K context: 215 GB8-bit, 8K context: 358 GB
Parameters
358B
Active per token
35.6B
Longest context
128K
Released
Jul 2025
License
MIT
Strongest at
Not rated

How much VRAM GLM 4.5 needs

Memory GLM 4.5 needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
16-bit667 GB671 GB680 GB714 GB
8-bit355 GB358 GB367 GB401 GB
6-bit274 GB277 GB286 GB321 GB
5-bit237 GB241 GB250 GB284 GB
4-bit202 GB206 GB215 GB249 GB
3-bit163 GB167 GB176 GB210 GB
2-bit125 GB129 GB138 GB172 GB
1.6-bit68.8 GB72.6 GB81.2 GB116 GB

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

GPUs and Macs that run GLM 4.5

The best way GLM 4.5 runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: GLM 4.5 on RTX 4060 report
RTX 5060Too bigPDF: GLM 4.5 on RTX 5060 report
RTX 3070Too bigPDF: GLM 4.5 on RTX 3070 report
RTX 3080 10 GBToo bigPDF: GLM 4.5 on RTX 3080 10 GB report
RTX 3060 12 GBToo bigPDF: GLM 4.5 on RTX 3060 12 GB report
Arc B580Too bigPDF: GLM 4.5 on Arc B580 report
RTX 4070Too bigPDF: GLM 4.5 on RTX 4070 report
RTX 4070 SuperToo bigPDF: GLM 4.5 on RTX 4070 Super report
RTX 5070Too bigPDF: GLM 4.5 on RTX 5070 report
RTX 4060 Ti 16 GBToo bigPDF: GLM 4.5 on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GBToo bigPDF: GLM 4.5 on RTX 5060 Ti 16 GB report
Arc A770 16 GBToo bigPDF: GLM 4.5 on Arc A770 16 GB report
Radeon RX 9070 XTToo bigPDF: GLM 4.5 on Radeon RX 9070 XT report
RTX 4070 Ti SuperToo bigPDF: GLM 4.5 on RTX 4070 Ti Super report
RTX 4080 SuperToo bigPDF: GLM 4.5 on RTX 4080 Super report
RTX 5070 TiToo bigPDF: GLM 4.5 on RTX 5070 Ti report
RTX 5080Too bigPDF: GLM 4.5 on RTX 5080 report
Radeon RX 7900 XTXToo bigPDF: GLM 4.5 on Radeon RX 7900 XTX report
RTX 3090Too bigPDF: GLM 4.5 on RTX 3090 report
RTX 4090Too bigPDF: GLM 4.5 on RTX 4090 report
Radeon AI PRO R9700Too bigPDF: GLM 4.5 on Radeon AI PRO R9700 report
RTX 5090Too bigPDF: GLM 4.5 on RTX 5090 report
2× RTX 3090Too bigPDF: GLM 4.5 on 2× RTX 3090 report
2× RTX 4090Too bigPDF: GLM 4.5 on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GBToo bigPDF: GLM 4.5 on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 AdaToo bigPDF: GLM 4.5 on RTX 6000 Ada report
L40SToo bigPDF: GLM 4.5 on L40S report
2× RTX 5090Too bigPDF: GLM 4.5 on 2× RTX 5090 report
4× RTX 3090Too bigPDF: GLM 4.5 on 4× RTX 3090 report
RTX PRO 6000 BlackwellToo bigPDF: GLM 4.5 on RTX PRO 6000 Blackwell report
Mac mini M4 16 GBToo bigPDF: GLM 4.5 on Mac mini M4 16 GB report
Mac mini M4 24 GBToo bigPDF: GLM 4.5 on Mac mini M4 24 GB report
Mac mini M4 32 GBToo bigPDF: GLM 4.5 on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GBToo bigPDF: GLM 4.5 on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GBToo bigPDF: GLM 4.5 on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GBToo bigPDF: GLM 4.5 on Mac mini M4 Pro 64 GB report
M4 Max 128 GBToo bigPDF: GLM 4.5 on M4 Max 128 GB report
M5 Max 128 GBToo bigPDF: GLM 4.5 on M5 Max 128 GB report
M3 Ultra 96 GBToo bigPDF: GLM 4.5 on M3 Ultra 96 GB report
M3 Ultra 256 GB3-bit48K22 tok/s22 sPDF: GLM 4.5 on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit24K12 tok/s22 sPDF: GLM 4.5 on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit24K16 tok/s7.0 sPDF: GLM 4.5 on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GBToo bigPDF: GLM 4.5 on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GBToo bigPDF: GLM 4.5 on Ryzen AI Max+ 395 128 GB report
RTX SparkToo bigPDF: GLM 4.5 on RTX Spark report
DGX SparkToo bigPDF: GLM 4.5 on DGX Spark report
Dual DGX Spark4-bit64K8.4 tok/s4.6 sPDF: GLM 4.5 on Dual DGX Spark report
DGX Station GB3004-bit96K101 tok/s0.6 sPDF: GLM 4.5 on DGX Station GB300 report
2× RTX PRO 6000 Blackwell3-bit48K52 tok/s1.1 sPDF: GLM 4.5 on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit20K28 tok/s0.6 sPDF: GLM 4.5 on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit128K28 tok/s0.3 sPDF: GLM 4.5 on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXMToo bigPDF: GLM 4.5 on A100 80 GB SXM report
H100 SXMToo bigPDF: GLM 4.5 on H100 SXM report
4× H200 NVL8-bit128K60 tok/s0.6 sPDF: GLM 4.5 on 4× H200 NVL report
8× H200 NVL8-bit128K60 tok/s0.3 sPDF: GLM 4.5 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-GGUF
vLLM
vllm serve zai-org/GLM-4.5
MLX (Mac)
mlx_lm.chat --model mlx-community/GLM-4.5-4bit

Questions

How much VRAM does GLM 4.5 need?

GLM 4.5 needs 202 GB for its weights at 4-bit, 206 GB with an 8K context and 215 GB with 32K. At 8-bit, 358 GB.

What is the minimum VRAM for GLM 4.5?

About 129 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 M3 Ultra 256 GB.

How big is GLM 4.5?

358B parameters (35.6B active per token): 667 GB of weights at 16-bit, 202 GB at 4-bit.

Can GLM 4.5 run on an RTX 4090?

Not on one card. The smallest machine here that runs it is the M3 Ultra 256 GB.

More models