How much VRAM does Laguna-M.1 need?

Q8_0 is exactly 240,119,891,264 bytes (223.63 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is Laguna-M.1's KV cache?

8.75 GiB at 32K context with an f16 cache, computed per layer from the model's architecture. Quantizing the cache to q8_0 roughly halves it.

Is Laguna-M.1 a mixture-of-experts model?

Yes — 256 experts with 16 routed per token. Every expert must be resident in memory, but only the routed ones are read per token, so its memory requirement and its speed behave very differently.

poolside · text · mixture of experts

Laguna-M.1

poolside/Laguna-M.1

Laguna-M.1 at Q8_0 is exactly 240,119,891,264 bytes (223.63 GiB / 240.12 GB) — an effective 8.508 bits per weight, not the nominal 8. Its KV cache at 32K is 8.75 GiB.

From the file· summed from 10 file(s)From the file· KV per layer

Parameters

226B

total, not active

Architecture

laguna

70 layers

Context

262,144

native (config.json)

License

apache-2.0

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
Q8_010 shards	223.63 GiB	240,119,891,264	8.508	—	sigargv
BF1610 shards	420.71 GiB	451,736,945,984	16.005	—	sigargv

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	1.09 GiB	1.09 GiB	—	70 / 0 / 0
8,192	2.19 GiB	2.19 GiB	—	70 / 0 / 0
16,384	4.38 GiB	4.38 GiB	—	70 / 0 / 0
32,768	8.75 GiB	8.75 GiB	—	70 / 0 / 0
65,536	17.50 GiB	17.50 GiB	—	70 / 0 / 0
131,072	35.00 GiB	35.00 GiB	—	70 / 0 / 0

Compare with

same modality, comparable size

vs DeepSeek-V4-Flash vs DeepSeek-V4-Flash-0731 vs Laguna-S-2.1 vs Hy3 vs GLM-5.2 vs Qwen3-Coder-Next All alternatives →Will it fit on my card? →

Why other calculators give a different number

A parameters × bits ÷ 8 estimate puts Q8_0 at roughly 118.29 GiB. The real file is 223.63 GiB, because a quantization is a mixture and some tensors are always kept at higher precision.

Architecture

from config.json

Layers

Attention heads

KV heads

Head dim

128

Hidden size

4096

Vocab

100,352

Sliding window

SWA period

MLA

Experts

256

Experts per token

use_sliding_window

—

Questions people ask

How much VRAM does Laguna-M.1 need?: Q8_0 is exactly 240,119,891,264 bytes (223.63 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is Laguna-M.1's KV cache?: 8.75 GiB at 32K context with an f16 cache, computed per layer. Quantizing the cache to q8_0 roughly halves it, which is often the difference between a context length fitting and not.
Is Laguna-M.1 a mixture-of-experts model?: Yes — 256 experts, 16 routed per token. Every expert must be resident, but only the routed ones are read per token, which is why its memory requirement and its speed behave very differently.
Which quantization of Laguna-M.1 should I use?: Q4_K_M is the usual default. Pick the largest quantization that fits your card at the context you actually need — the table above gives exact sizes for every one published.