Is EXAONE-4.0-1.2B-abliterated or Qwen3-1.7B smaller?

EXAONE-4.0-1.2B-abliterated is smaller at Q4_K_M: 0.87 GiB against 1.03 GiB.

Which uses less memory at long context?

EXAONE-4.0-1.2B-abliterated — 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 Qwen3-1.7B

At Q4_K_M, EXAONE-4.0-1.2B-abliterated is the smaller download — 930,403,456 bytes against 1,107,409,472. At long context the gap widens: EXAONE-4.0-1.2B-abliterated'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	Qwen3-1.7B
Parameters	1.5B	2.0B
Architecture	exaone4	qwen3
Layers	30	28
Native context	65,536	40,960
Mixture of experts	no	no
Quantizations published	36	48
Smallest quantization	0.42 GiB	0.50 GiB
Q4_K_M	0.87 GiB	1.03 GiB
Licence	—	—

KV cache by context

the term that decides long-context viability

Context	EXAONE-4.0-1.2B-abliterated	Qwen3-1.7B	Ratio
4,096	0.23 GiB	0.44 GiB	1.87×
8,192	0.47 GiB	0.88 GiB	1.87×
16,384	0.94 GiB	1.75 GiB	1.87×
32,768	1.88 GiB	3.50 GiB	1.87×
65,536	3.75 GiB	7.00 GiB	1.87×
131,072	7.50 GiB	14.00 GiB	1.87×