Is LFM2-2.6B-Exp or Llama-3.2-1B-Instruct smaller?

Llama-3.2-1B-Instruct is smaller at Q4_K_M: 0.75 GiB against 1.46 GiB.

Which uses less memory at long context?

LFM2-2.6B-Exp — its KV cache at 32K context is 2.0x smaller, because of how its attention layers cache.

Model comparison

LFM2-2.6B-Exp vs Llama-3.2-1B-Instruct

At Q4_K_M, Llama-3.2-1B-Instruct is the smaller download — 807,690,688 bytes against 1,563,669,120. At long context the gap widens: LFM2-2.6B-Exp's KV cache at 32K is 2.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	LFM2-2.6B-Exp	Llama-3.2-1B-Instruct
Parameters	2.6B	1.2B
Architecture	lfm2	llama
Layers	30	16
Native context	128,000	131,072
Mixture of experts	no	no
Quantizations published	42	39
Smallest quantization	0.88 GiB	0.39 GiB
Q4_K_M	1.46 GiB	0.75 GiB
Licence	other	—

KV cache by context

the term that decides long-context viability

Context	LFM2-2.6B-Exp	Llama-3.2-1B-Instruct	Ratio
4,096	0.06 GiB	0.13 GiB	2.00×
8,192	0.13 GiB	0.25 GiB	2.00×
16,384	0.25 GiB	0.50 GiB	2.00×
32,768	0.50 GiB	1.00 GiB	2.00×
65,536	1.00 GiB	2.00 GiB	2.00×
131,072	2.00 GiB	4.00 GiB	2.00×