Skip to content

Laguna M.1 VRAM requirements

How much VRAM Laguna M.1 needs
4-bit weights: 127 GBWith 8K context: 130 GBWith 32K context: 137 GB8-bit, 8K context: 226 GB
Parameters
226B
Active per token
23.5B
Longest context
256K
Released
Jun 2026
License
Apache 2.0
Strongest at
Not rated

How much VRAM Laguna M.1 needs

Memory Laguna M.1 needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
16-bit421 GB424 GB430 GB456 GB
8-bit223 GB226 GB233 GB259 GB
6-bit172 GB175 GB182 GB208 GB
5-bit150 GB153 GB159 GB185 GB
4-bit127 GB130 GB137 GB163 GB
3-bit103 GB106 GB112 GB139 GB
2-bit78.9 GB81.8 GB88.4 GB115 GB
1.6-bit43.4 GB46.3 GB52.9 GB79.1 GB

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

GPUs and Macs that run Laguna M.1

The best way Laguna M.1 runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: Laguna M.1 on RTX 4060 report
RTX 5060Too bigPDF: Laguna M.1 on RTX 5060 report
RTX 3070Too bigPDF: Laguna M.1 on RTX 3070 report
RTX 3080 10 GBToo bigPDF: Laguna M.1 on RTX 3080 10 GB report
RTX 3060 12 GBToo bigPDF: Laguna M.1 on RTX 3060 12 GB report
Arc B580Too bigPDF: Laguna M.1 on Arc B580 report
RTX 4070Too bigPDF: Laguna M.1 on RTX 4070 report
RTX 4070 SuperToo bigPDF: Laguna M.1 on RTX 4070 Super report
RTX 5070Too bigPDF: Laguna M.1 on RTX 5070 report
RTX 4060 Ti 16 GBToo bigPDF: Laguna M.1 on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GBToo bigPDF: Laguna M.1 on RTX 5060 Ti 16 GB report
Arc A770 16 GBToo bigPDF: Laguna M.1 on Arc A770 16 GB report
Radeon RX 9070 XTToo bigPDF: Laguna M.1 on Radeon RX 9070 XT report
RTX 4070 Ti SuperToo bigPDF: Laguna M.1 on RTX 4070 Ti Super report
RTX 4080 SuperToo bigPDF: Laguna M.1 on RTX 4080 Super report
RTX 5070 TiToo bigPDF: Laguna M.1 on RTX 5070 Ti report
RTX 5080Too bigPDF: Laguna M.1 on RTX 5080 report
Radeon RX 7900 XTXToo bigPDF: Laguna M.1 on Radeon RX 7900 XTX report
RTX 3090Too bigPDF: Laguna M.1 on RTX 3090 report
RTX 4090Too bigPDF: Laguna M.1 on RTX 4090 report
Radeon AI PRO R9700Too bigPDF: Laguna M.1 on Radeon AI PRO R9700 report
RTX 5090Too bigPDF: Laguna M.1 on RTX 5090 report
2× RTX 3090Too bigPDF: Laguna M.1 on 2× RTX 3090 report
2× RTX 4090Too bigPDF: Laguna M.1 on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GBToo bigPDF: Laguna M.1 on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 AdaToo bigPDF: Laguna M.1 on RTX 6000 Ada report
L40SToo bigPDF: Laguna M.1 on L40S report
2× RTX 5090Too bigPDF: Laguna M.1 on 2× RTX 5090 report
4× RTX 30902-bit32K57 tok/s1.4 sPDF: Laguna M.1 on 4× RTX 3090 report
RTX PRO 6000 Blackwell2-bit40K89 tok/s1.3 sPDF: Laguna M.1 on RTX PRO 6000 Blackwell report
Mac mini M4 16 GBToo bigPDF: Laguna M.1 on Mac mini M4 16 GB report
Mac mini M4 24 GBToo bigPDF: Laguna M.1 on Mac mini M4 24 GB report
Mac mini M4 32 GBToo bigPDF: Laguna M.1 on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GBToo bigPDF: Laguna M.1 on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GBToo bigPDF: Laguna M.1 on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GBToo bigPDF: Laguna M.1 on Mac mini M4 Pro 64 GB report
M4 Max 128 GB2-bit40K32 tok/s27 sPDF: Laguna M.1 on M4 Max 128 GB report
M5 Max 128 GB2-bit40K35 tok/s7.7 sPDF: Laguna M.1 on M5 Max 128 GB report
M3 Ultra 96 GBToo bigPDF: Laguna M.1 on M3 Ultra 96 GB report
M3 Ultra 256 GB6-bit32K22 tok/s15 sPDF: Laguna M.1 on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit256K18 tok/s15 sPDF: Laguna M.1 on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit256K24 tok/s4.7 sPDF: Laguna M.1 on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GBToo bigPDF: Laguna M.1 on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB2-bit40K16 tok/s15 sPDF: Laguna M.1 on Ryzen AI Max+ 395 128 GB report
RTX Spark3-bit36K17 tok/s5.3 sPDF: Laguna M.1 on RTX Spark report
DGX Spark3-bit36K16 tok/s5.3 sPDF: Laguna M.1 on DGX Spark report
Dual DGX Spark6-bit200K9.5 tok/s3.1 sPDF: Laguna M.1 on Dual DGX Spark report
DGX Station GB3008-bit52K109 tok/s0.4 sPDF: Laguna M.1 on DGX Station GB300 report
2× RTX PRO 6000 Blackwell6-bit28K51 tok/s0.7 sPDF: Laguna M.1 on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit256K42 tok/s0.4 sPDF: Laguna M.1 on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit256K42 tok/s0.2 sPDF: Laguna M.1 on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXMToo bigPDF: Laguna M.1 on A100 80 GB SXM report
H100 SXMToo bigPDF: Laguna M.1 on H100 SXM report
4× H200 NVL8-bit256K86 tok/s0.4 sPDF: Laguna M.1 on 4× H200 NVL report
8× H200 NVL8-bit256K86 tok/s0.2 sPDF: Laguna M.1 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 sigargv/Laguna-M.1-GGUF
vLLM
vllm serve poolside/Laguna-M.1
MLX (Mac)
mlx_lm.chat --model mlx-community/Laguna-M.1-4bit

Questions

How much VRAM does Laguna M.1 need?

Laguna M.1 needs 127 GB for its weights at 4-bit, 130 GB with an 8K context and 137 GB with 32K. At 8-bit, 226 GB.

What is the minimum VRAM for Laguna M.1?

About 81.8 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 4× RTX 3090.

How big is Laguna M.1?

226B parameters (23.5B active per token): 421 GB of weights at 16-bit, 127 GB at 4-bit.

Can Laguna M.1 run on an RTX 4090?

Not on one card. The smallest machine here that runs it is the 4× RTX 3090.

More models