Skip to content

GLM 4.6 VRAM requirements

How much VRAM GLM 4.6 needs
4-bit weights: 201 GBWith 8K context: 205 GBWith 32K context: 214 GB8-bit, 8K context: 357 GB
Parameters
357B
Active per token
34B
Longest context
198K
Released
Sep 2025
License
MIT
Strongest at
Not rated

How much VRAM GLM 4.6 needs

Memory GLM 4.6 needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
16-bit665 GB668 GB677 GB711 GB
8-bit353 GB357 GB365 GB400 GB
6-bit272 GB276 GB285 GB319 GB
5-bit236 GB240 GB249 GB283 GB
4-bit201 GB205 GB214 GB248 GB
3-bit162 GB166 GB175 GB209 GB
2-bit125 GB128 GB137 GB172 GB
1.6-bit68.5 GB72.3 GB80.9 GB115 GB

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

GPUs and Macs that run GLM 4.6

The best way GLM 4.6 runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: GLM 4.6 on RTX 4060 report
RTX 5060Too bigPDF: GLM 4.6 on RTX 5060 report
RTX 3070Too bigPDF: GLM 4.6 on RTX 3070 report
RTX 3080 10 GBToo bigPDF: GLM 4.6 on RTX 3080 10 GB report
RTX 3060 12 GBToo bigPDF: GLM 4.6 on RTX 3060 12 GB report
Arc B580Too bigPDF: GLM 4.6 on Arc B580 report
RTX 4070Too bigPDF: GLM 4.6 on RTX 4070 report
RTX 4070 SuperToo bigPDF: GLM 4.6 on RTX 4070 Super report
RTX 5070Too bigPDF: GLM 4.6 on RTX 5070 report
RTX 4060 Ti 16 GBToo bigPDF: GLM 4.6 on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GBToo bigPDF: GLM 4.6 on RTX 5060 Ti 16 GB report
Arc A770 16 GBToo bigPDF: GLM 4.6 on Arc A770 16 GB report
Radeon RX 9070 XTToo bigPDF: GLM 4.6 on Radeon RX 9070 XT report
RTX 4070 Ti SuperToo bigPDF: GLM 4.6 on RTX 4070 Ti Super report
RTX 4080 SuperToo bigPDF: GLM 4.6 on RTX 4080 Super report
RTX 5070 TiToo bigPDF: GLM 4.6 on RTX 5070 Ti report
RTX 5080Too bigPDF: GLM 4.6 on RTX 5080 report
Radeon RX 7900 XTXToo bigPDF: GLM 4.6 on Radeon RX 7900 XTX report
RTX 3090Too bigPDF: GLM 4.6 on RTX 3090 report
RTX 4090Too bigPDF: GLM 4.6 on RTX 4090 report
Radeon AI PRO R9700Too bigPDF: GLM 4.6 on Radeon AI PRO R9700 report
RTX 5090Too bigPDF: GLM 4.6 on RTX 5090 report
2× RTX 3090Too bigPDF: GLM 4.6 on 2× RTX 3090 report
2× RTX 4090Too bigPDF: GLM 4.6 on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GBToo bigPDF: GLM 4.6 on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 AdaToo bigPDF: GLM 4.6 on RTX 6000 Ada report
L40SToo bigPDF: GLM 4.6 on L40S report
2× RTX 5090Too bigPDF: GLM 4.6 on 2× RTX 5090 report
4× RTX 3090Too bigPDF: GLM 4.6 on 4× RTX 3090 report
RTX PRO 6000 BlackwellToo bigPDF: GLM 4.6 on RTX PRO 6000 Blackwell report
Mac mini M4 16 GBToo bigPDF: GLM 4.6 on Mac mini M4 16 GB report
Mac mini M4 24 GBToo bigPDF: GLM 4.6 on Mac mini M4 24 GB report
Mac mini M4 32 GBToo bigPDF: GLM 4.6 on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GBToo bigPDF: GLM 4.6 on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GBToo bigPDF: GLM 4.6 on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GBToo bigPDF: GLM 4.6 on Mac mini M4 Pro 64 GB report
M4 Max 128 GBToo bigPDF: GLM 4.6 on M4 Max 128 GB report
M5 Max 128 GBToo bigPDF: GLM 4.6 on M5 Max 128 GB report
M3 Ultra 96 GBToo bigPDF: GLM 4.6 on M3 Ultra 96 GB report
M3 Ultra 256 GB3-bit52K23 tok/s21 sPDF: GLM 4.6 on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit28K12 tok/s21 sPDF: GLM 4.6 on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit28K17 tok/s6.7 sPDF: GLM 4.6 on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GBToo bigPDF: GLM 4.6 on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GBToo bigPDF: GLM 4.6 on Ryzen AI Max+ 395 128 GB report
RTX SparkToo bigPDF: GLM 4.6 on RTX Spark report
DGX SparkToo bigPDF: GLM 4.6 on DGX Spark report
Dual DGX Spark4-bit72K8.8 tok/s4.4 sPDF: GLM 4.6 on Dual DGX Spark report
DGX Station GB3004-bit96K103 tok/s0.6 sPDF: GLM 4.6 on DGX Station GB300 report
2× RTX PRO 6000 Blackwell3-bit48K53 tok/s1.0 sPDF: GLM 4.6 on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit24K29 tok/s0.6 sPDF: GLM 4.6 on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit198K29 tok/s0.3 sPDF: GLM 4.6 on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXMToo bigPDF: GLM 4.6 on A100 80 GB SXM report
H100 SXMToo bigPDF: GLM 4.6 on H100 SXM report
4× H200 NVL8-bit198K61 tok/s0.6 sPDF: GLM 4.6 on 4× H200 NVL report
8× H200 NVL8-bit198K61 tok/s0.3 sPDF: GLM 4.6 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-4.6-GGUF
vLLM
vllm serve zai-org/GLM-4.6
MLX (Mac)
mlx_lm.chat --model mlx-community/GLM-4.6-4bit

Questions

How much VRAM does GLM 4.6 need?

GLM 4.6 needs 201 GB for its weights at 4-bit, 205 GB with an 8K context and 214 GB with 32K. At 8-bit, 357 GB.

What is the minimum VRAM for GLM 4.6?

About 128 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 4.6?

357B parameters (34B active per token): 665 GB of weights at 16-bit, 201 GB at 4-bit.

Can GLM 4.6 run on an RTX 4090?

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

More models