Skip to content

GLM-5.3 VRAM requirements

How much VRAM GLM-5.3 needs
4-bit weights: 425 GBWith 8K context: 427 GBWith 32K context: 429 GB
Parameters
753B
Active per token
40B
Longest context
1M
Released
Aug 2026
License
GLM-5.3 License
Strongest at
Chat, Coding, Agents, Docs

How much VRAM GLM-5.3 needs

Memory GLM-5.3 needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
As released (FP8)704 GB705 GB708 GB717 GB
6-bit575 GB577 GB579 GB588 GB
5-bit499 GB501 GB503 GB512 GB
4-bit425 GB427 GB429 GB438 GB
3-bit343 GB345 GB347 GB356 GB
2-bit263 GB265 GB267 GB276 GB
1.6-bit145 GB146 GB149 GB158 GB

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

GPUs and Macs that run GLM-5.3

The best way GLM-5.3 runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: GLM-5.3 on RTX 4060 report
RTX 5060Too bigPDF: GLM-5.3 on RTX 5060 report
RTX 3070Too bigPDF: GLM-5.3 on RTX 3070 report
RTX 3080 10 GBToo bigPDF: GLM-5.3 on RTX 3080 10 GB report
RTX 3060 12 GBToo bigPDF: GLM-5.3 on RTX 3060 12 GB report
Arc B580Too bigPDF: GLM-5.3 on Arc B580 report
RTX 4070Too bigPDF: GLM-5.3 on RTX 4070 report
RTX 4070 SuperToo bigPDF: GLM-5.3 on RTX 4070 Super report
RTX 5070Too bigPDF: GLM-5.3 on RTX 5070 report
RTX 4060 Ti 16 GBToo bigPDF: GLM-5.3 on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GBToo bigPDF: GLM-5.3 on RTX 5060 Ti 16 GB report
Arc A770 16 GBToo bigPDF: GLM-5.3 on Arc A770 16 GB report
Radeon RX 9070 XTToo bigPDF: GLM-5.3 on Radeon RX 9070 XT report
RTX 4070 Ti SuperToo bigPDF: GLM-5.3 on RTX 4070 Ti Super report
RTX 4080 SuperToo bigPDF: GLM-5.3 on RTX 4080 Super report
RTX 5070 TiToo bigPDF: GLM-5.3 on RTX 5070 Ti report
RTX 5080Too bigPDF: GLM-5.3 on RTX 5080 report
Radeon RX 7900 XTXToo bigPDF: GLM-5.3 on Radeon RX 7900 XTX report
RTX 3090Too bigPDF: GLM-5.3 on RTX 3090 report
RTX 4090Too bigPDF: GLM-5.3 on RTX 4090 report
Radeon AI PRO R9700Too bigPDF: GLM-5.3 on Radeon AI PRO R9700 report
RTX 5090Too bigPDF: GLM-5.3 on RTX 5090 report
2× RTX 3090Too bigPDF: GLM-5.3 on 2× RTX 3090 report
2× RTX 4090Too bigPDF: GLM-5.3 on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GBToo bigPDF: GLM-5.3 on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 AdaToo bigPDF: GLM-5.3 on RTX 6000 Ada report
L40SToo bigPDF: GLM-5.3 on L40S report
2× RTX 5090Too bigPDF: GLM-5.3 on 2× RTX 5090 report
4× RTX 3090Too bigPDF: GLM-5.3 on 4× RTX 3090 report
RTX PRO 6000 BlackwellToo bigPDF: GLM-5.3 on RTX PRO 6000 Blackwell report
Mac mini M4 16 GBToo bigPDF: GLM-5.3 on Mac mini M4 16 GB report
Mac mini M4 24 GBToo bigPDF: GLM-5.3 on Mac mini M4 24 GB report
Mac mini M4 32 GBToo bigPDF: GLM-5.3 on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GBToo bigPDF: GLM-5.3 on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GBToo bigPDF: GLM-5.3 on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GBToo bigPDF: GLM-5.3 on Mac mini M4 Pro 64 GB report
M4 Max 128 GBToo bigPDF: GLM-5.3 on M4 Max 128 GB report
M5 Max 128 GBToo bigPDF: GLM-5.3 on M5 Max 128 GB report
M3 Ultra 96 GBToo bigPDF: GLM-5.3 on M3 Ultra 96 GB report
M3 Ultra 256 GB1.6-bit384K38 tok/s25 sPDF: GLM-5.3 on M3 Ultra 256 GB report
M3 Ultra 512 GB3-bit216K21 tok/s25 sPDF: GLM-5.3 on M3 Ultra 512 GB report
M5 Ultra 512 GB3-bit216K28 tok/s7.8 sPDF: GLM-5.3 on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GBToo bigPDF: GLM-5.3 on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GBToo bigPDF: GLM-5.3 on Ryzen AI Max+ 395 128 GB report
RTX SparkToo bigPDF: GLM-5.3 on RTX Spark report
DGX SparkToo bigPDF: GLM-5.3 on DGX Spark report
Dual DGX Spark1.6-bit864K21 tok/s5.1 sPDF: GLM-5.3 on Dual DGX Spark report
DGX Station GB3001.6-bit960K158 tok/s0.7 sPDF: GLM-5.3 on DGX Station GB300 report
2× RTX PRO 6000 Blackwell1.6-bit352K89 tok/s1.2 sPDF: GLM-5.3 on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell3-bit192K49 tok/s0.7 sPDF: GLM-5.3 on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 BlackwellFP8208K27 tok/s0.4 sPDF: GLM-5.3 on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXMToo bigPDF: GLM-5.3 on A100 80 GB SXM report
H100 SXMToo bigPDF: GLM-5.3 on H100 SXM report
4× H200 NVL5-bit352K76 tok/s0.7 sPDF: GLM-5.3 on 4× H200 NVL report
8× H200 NVLFP81M60 tok/s0.3 sPDF: GLM-5.3 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-5.3-GGUF
vLLM
vllm serve zai-org/GLM-5.3
MLX (Mac)
mlx_lm.chat --model mlx-community/GLM-5.3-4bit

Questions

How much VRAM does GLM-5.3 need?

GLM-5.3 needs 425 GB for its weights at 4-bit, 427 GB with an 8K context and 429 GB with 32K.

What is the minimum VRAM for GLM-5.3?

About 146 GB at 1.6-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-5.3?

753B parameters (40B active per token): 704 GB of weights as released (FP8), 425 GB at 4-bit.

Can GLM-5.3 run on an RTX 4090?

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

More models