Is LFM-2.5-1.2b-Instruct-roleplay-tuned or gemma-3-1b-it smaller?

LFM-2.5-1.2b-Instruct-roleplay-tuned is smaller at Q4_K_M: 0.68 GiB against 0.75 GiB.

Which uses less memory at long context?

gemma-3-1b-it — its KV cache at 32K context is 2.6x smaller, because of how its attention layers cache.

Model comparison

LFM-2.5-1.2b-Instruct-roleplay-tuned vs gemma-3-1b-it

At Q4_K_M, LFM-2.5-1.2b-Instruct-roleplay-tuned is the smaller download — 730,896,288 bytes against 806,058,240. At long context the gap widens: gemma-3-1b-it's KV cache at 32K is 2.6× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	LFM-2.5-1.2b-Instruct-roleplay-tuned	gemma-3-1b-it
Parameters	1.2B	1000M
Architecture	lfm2	gemma3
Layers	16	26
Native context	128,000	32,768
Mixture of experts	no	no
Quantizations published	36	28
Smallest quantization	0.28 GiB	0.52 GiB
Q4_K_M	0.68 GiB	0.75 GiB
Licence	other	gemma

KV cache by context

the term that decides long-context viability

Context	LFM-2.5-1.2b-Instruct-roleplay-tuned	gemma-3-1b-it	Ratio
4,096	0.05 GiB	0.04 GiB	1.26×
8,192	0.09 GiB	0.05 GiB	1.78×
16,384	0.19 GiB	0.08 GiB	2.23×
32,768	0.38 GiB	0.15 GiB	2.56×
65,536	0.75 GiB	0.27 GiB	2.76×
131,072	1.50 GiB	0.52 GiB	2.88×