Is bitnet_b1_58-large or LFM2.5-1.2B-Instruct smaller?

Both publish quantizations; compare the table for the exact sizes.

Which uses less memory at long context?

LFM2.5-1.2B-Instruct — its KV cache at 32K context is 12.0x smaller, because of how its attention layers cache.

Model comparison

bitnet_b1_58-large vs LFM2.5-1.2B-Instruct

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: LFM2.5-1.2B-Instruct's KV cache at 32K is 12.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	bitnet_b1_58-large	LFM2.5-1.2B-Instruct
Parameters	729M	1.2B
Architecture	bitnet	lfm2
Layers	24	16
Native context	2,048	128,000
Mixture of experts	no	no
Quantizations published	1	23
Smallest quantization	0.20 GiB	0.45 GiB
Q4_K_M	—	0.68 GiB
Licence	mit	other

KV cache by context

the term that decides long-context viability

Context	bitnet_b1_58-large	LFM2.5-1.2B-Instruct	Ratio
4,096	0.56 GiB	0.05 GiB	12.00×
8,192	1.13 GiB	0.09 GiB	12.00×
16,384	2.25 GiB	0.19 GiB	12.00×
32,768	4.50 GiB	0.38 GiB	12.00×
65,536	9.00 GiB	0.75 GiB	12.00×
131,072	18.00 GiB	1.50 GiB	12.00×