How much VRAM does LFM2.5-350M need?

Q4_K_M is exactly 228,332,128 bytes (0.21 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is LFM2.5-350M's KV cache?

0.38 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.

LiquidAI · text

LFM2.5-350M

LiquidAI/LFM2.5-350M

LFM2.5-350M at Q4_K_M is exactly 228,332,128 bytes (0.21 GiB / 0.23 GB) — an effective 5.153 bits per weight, not the nominal 4. Its KV cache at 32K is 0.38 GiB.

From the file· summed from 1 file(s)From the file· KV per layer

Parameters

354M

Architecture

lfm2

16 layers

Context

128,000

native (config.json)

License

other

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
Q4_0	0.20 GiB	219,309,792	4.949	—	LiquidAI
Q4_K_M	0.21 GiB	228,332,128	5.153	—	OrdenWills
Q4_K_M	0.21 GiB	229,312,224	5.175	—	LiquidAI
Q4_K_M	0.21 GiB	229,314,080	5.175	148	lazos
Q5_K_M	0.24 GiB	260,376,288	5.876	—	LiquidAI
Q6_K	0.27 GiB	293,381,856	6.621	—	LiquidAI
Q6_K	0.27 GiB	293,383,712	6.621	148	lazos
Q8_0	0.35 GiB	377,956,288	8.530	—	OrdenWills
Q8_0	0.35 GiB	379,217,632	8.558	—	LiquidAI
Q8_0	0.35 GiB	379,219,488	8.558	148	lazos
F16	0.66 GiB	709,135,168	16.004	—	OrdenWills
F16	0.66 GiB	711,485,152	16.057	—	LiquidAI
BF16	0.66 GiB	711,485,152	16.057	—	LiquidAI

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	0.05 GiB	0.13 GiB	2.67×	6 / 0 / 10
8,192	0.09 GiB	0.25 GiB	2.67×	6 / 0 / 10
16,384	0.19 GiB	0.50 GiB	2.67×	6 / 0 / 10
32,768	0.38 GiB	1.00 GiB	2.67×	6 / 0 / 10
65,536	0.75 GiB	2.00 GiB	2.67×	6 / 0 / 10
131,072	1.50 GiB	4.00 GiB	2.67×	6 / 0 / 10

10 of 16 layers use linear attention, which keeps a fixed-size recurrent state instead of a per-token cache. Those layers do not grow with context at all — treating them as ordinary attention, as a flat formula does, overstates this model's cache by roughly 2.7× at long context.

Compare with

same modality, comparable size

vs Llama-3.2-1B-Instruct vs ced-base vs Qwen3-1.7B vs embeddinggemma-300m vs Qwen3-0.6B vs LFM2.5-1.2B-Instruct All alternatives →Will it fit on my card? →

Why other calculators give a different number

A parameters × bits ÷ 8 estimate puts Q4_K_M at roughly 0.19 GiB. The real file is 0.21 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

Hidden size

1024

Vocab

65,536

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

—

Questions people ask

How much VRAM does LFM2.5-350M need?: Q4_K_M is exactly 228,332,128 bytes (0.21 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is LFM2.5-350M's KV cache?: 0.38 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.
Which quantization of LFM2.5-350M 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.