Is EXAONE-4.0-1.2B-abliterated or Llama-3.2-1B-Instruct smaller?

Llama-3.2-1B-Instruct is smaller at Q4_K_M: 0.75 GiB against 0.87 GiB.

Which uses less memory at long context?

Llama-3.2-1B-Instruct — its KV cache at 32K context is 1.9x smaller, because of how its attention layers cache.

Model comparison

EXAONE-4.0-1.2B-abliterated vs Llama-3.2-1B-Instruct

At Q4_K_M, Llama-3.2-1B-Instruct is the smaller download — 807,690,688 bytes against 930,403,456. At long context the gap widens: Llama-3.2-1B-Instruct's KV cache at 32K is 1.9× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	EXAONE-4.0-1.2B-abliterated	Llama-3.2-1B-Instruct
Parameters	1.5B	1.2B
Architecture	exaone4	llama
Layers	30	16
Native context	65,536	131,072
Mixture of experts	no	no
Quantizations published	36	39
Smallest quantization	0.42 GiB	0.39 GiB
Q4_K_M	0.87 GiB	0.75 GiB
Licence	—	—

KV cache by context

the term that decides long-context viability

Context	EXAONE-4.0-1.2B-abliterated	Llama-3.2-1B-Instruct	Ratio
4,096	0.23 GiB	0.13 GiB	1.88×
8,192	0.47 GiB	0.25 GiB	1.88×
16,384	0.94 GiB	0.50 GiB	1.88×
32,768	1.88 GiB	1.00 GiB	1.88×
65,536	3.75 GiB	2.00 GiB	1.88×
131,072	7.50 GiB	4.00 GiB	1.88×