Skip to content

GLM 5.3 BF16 VRAM requirements

How much VRAM GLM 5.3 BF16 needs
4-bit weights: 425 GBWith 8K context: 427 GBWith 32K context: 429 GB8-bit, 8K context: 747 GB
Parameters
753B
Active per token
41.8B
Longest context
1M
Released
Aug 2026
License
GLM 5.3
Strongest at
Not rated

Its licence (GLM 5.3) is not a standard open-source one: check the terms before commercial use.

How much VRAM GLM 5.3 BF16 needs

Memory GLM 5.3 BF16 needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
16-bit1,403 GB1,405 GB1,407 GB1,416 GB
8-bit745 GB747 GB749 GB759 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 BF16

The best way GLM 5.3 BF16 runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: GLM 5.3 BF16 on RTX 4060 report
RTX 5060Too bigPDF: GLM 5.3 BF16 on RTX 5060 report
RTX 3070Too bigPDF: GLM 5.3 BF16 on RTX 3070 report
RTX 3080 10 GBToo bigPDF: GLM 5.3 BF16 on RTX 3080 10 GB report
RTX 3060 12 GBToo bigPDF: GLM 5.3 BF16 on RTX 3060 12 GB report
Arc B580Too bigPDF: GLM 5.3 BF16 on Arc B580 report
RTX 4070Too bigPDF: GLM 5.3 BF16 on RTX 4070 report
RTX 4070 SuperToo bigPDF: GLM 5.3 BF16 on RTX 4070 Super report
RTX 5070Too bigPDF: GLM 5.3 BF16 on RTX 5070 report
RTX 4060 Ti 16 GBToo bigPDF: GLM 5.3 BF16 on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GBToo bigPDF: GLM 5.3 BF16 on RTX 5060 Ti 16 GB report
Arc A770 16 GBToo bigPDF: GLM 5.3 BF16 on Arc A770 16 GB report
Radeon RX 9070 XTToo bigPDF: GLM 5.3 BF16 on Radeon RX 9070 XT report
RTX 4070 Ti SuperToo bigPDF: GLM 5.3 BF16 on RTX 4070 Ti Super report
RTX 4080 SuperToo bigPDF: GLM 5.3 BF16 on RTX 4080 Super report
RTX 5070 TiToo bigPDF: GLM 5.3 BF16 on RTX 5070 Ti report
RTX 5080Too bigPDF: GLM 5.3 BF16 on RTX 5080 report
Radeon RX 7900 XTXToo bigPDF: GLM 5.3 BF16 on Radeon RX 7900 XTX report
RTX 3090Too bigPDF: GLM 5.3 BF16 on RTX 3090 report
RTX 4090Too bigPDF: GLM 5.3 BF16 on RTX 4090 report
Radeon AI PRO R9700Too bigPDF: GLM 5.3 BF16 on Radeon AI PRO R9700 report
RTX 5090Too bigPDF: GLM 5.3 BF16 on RTX 5090 report
2× RTX 3090Too bigPDF: GLM 5.3 BF16 on 2× RTX 3090 report
2× RTX 4090Too bigPDF: GLM 5.3 BF16 on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GBToo bigPDF: GLM 5.3 BF16 on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 AdaToo bigPDF: GLM 5.3 BF16 on RTX 6000 Ada report
L40SToo bigPDF: GLM 5.3 BF16 on L40S report
2× RTX 5090Too bigPDF: GLM 5.3 BF16 on 2× RTX 5090 report
4× RTX 3090Too bigPDF: GLM 5.3 BF16 on 4× RTX 3090 report
RTX PRO 6000 BlackwellToo bigPDF: GLM 5.3 BF16 on RTX PRO 6000 Blackwell report
Mac mini M4 16 GBToo bigPDF: GLM 5.3 BF16 on Mac mini M4 16 GB report
Mac mini M4 24 GBToo bigPDF: GLM 5.3 BF16 on Mac mini M4 24 GB report
Mac mini M4 32 GBToo bigPDF: GLM 5.3 BF16 on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GBToo bigPDF: GLM 5.3 BF16 on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GBToo bigPDF: GLM 5.3 BF16 on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GBToo bigPDF: GLM 5.3 BF16 on Mac mini M4 Pro 64 GB report
M4 Max 128 GBToo bigPDF: GLM 5.3 BF16 on M4 Max 128 GB report
M5 Max 128 GBToo bigPDF: GLM 5.3 BF16 on M5 Max 128 GB report
M3 Ultra 96 GBToo bigPDF: GLM 5.3 BF16 on M3 Ultra 96 GB report
M3 Ultra 256 GB1.6-bit384K37 tok/s26 sPDF: GLM 5.3 BF16 on M3 Ultra 256 GB report
M3 Ultra 512 GB3-bit216K20 tok/s26 sPDF: GLM 5.3 BF16 on M3 Ultra 512 GB report
M5 Ultra 512 GB3-bit216K27 tok/s8.2 sPDF: GLM 5.3 BF16 on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GBToo bigPDF: GLM 5.3 BF16 on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GBToo bigPDF: GLM 5.3 BF16 on Ryzen AI Max+ 395 128 GB report
RTX SparkToo bigPDF: GLM 5.3 BF16 on RTX Spark report
DGX SparkToo bigPDF: GLM 5.3 BF16 on DGX Spark report
Dual DGX Spark1.6-bit864K20 tok/s5.4 sPDF: GLM 5.3 BF16 on Dual DGX Spark report
DGX Station GB3001.6-bit960K156 tok/s0.7 sPDF: GLM 5.3 BF16 on DGX Station GB300 report
2× RTX PRO 6000 Blackwell1.6-bit352K87 tok/s1.3 sPDF: GLM 5.3 BF16 on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell3-bit192K48 tok/s0.7 sPDF: GLM 5.3 BF16 on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell6-bit1M31 tok/s0.4 sPDF: GLM 5.3 BF16 on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXMToo bigPDF: GLM 5.3 BF16 on A100 80 GB SXM report
H100 SXMToo bigPDF: GLM 5.3 BF16 on H100 SXM report
4× H200 NVL5-bit352K74 tok/s0.7 sPDF: GLM 5.3 BF16 on 4× H200 NVL report
8× H200 NVL8-bit1M56 tok/s0.4 sPDF: GLM 5.3 BF16 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 AesSedai/GLM-5.3-GGUF
vLLM
vllm serve zai-org/GLM-5.3-BF16
MLX (Mac)
mlx_lm.chat --model mlx-community/GLM-5.3-4bit

Questions

How much VRAM does GLM 5.3 BF16 need?

GLM 5.3 BF16 needs 425 GB for its weights at 4-bit, 427 GB with an 8K context and 429 GB with 32K. At 8-bit, 747 GB.

What is the minimum VRAM for GLM 5.3 BF16?

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 BF16?

753B parameters (41.8B active per token): 1,403 GB of weights at 16-bit, 425 GB at 4-bit.

Can GLM 5.3 BF16 run on an RTX 4090?

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

More models