Is MARTHA-LXVII.8B_QWEN-3.5-3.6_prune_9b-3.6_base or Qwen3-4B smaller?

Qwen3-4B is smaller at Q4_K_M: 2.33 GiB against 4.76 GiB.

Which uses less memory at long context?

MARTHA-LXVII.8B_QWEN-3.5-3.6_prune_9b-3.6_base — its KV cache at 32K context is 5.1x smaller, because of how its attention layers cache.

Model comparison

MARTHA-LXVII.8B_QWEN-3.5-3.6_prune_9b-3.6_base vs Qwen3-4B

At Q4_K_M, Qwen3-4B is the smaller download — 2,497,280,256 bytes against 5,106,435,872. At long context the gap widens: MARTHA-LXVII.8B_QWEN-3.5-3.6_prune_9b-3.6_base's KV cache at 32K is 5.1× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	MARTHA-LXVII.8B_QWEN-3.5-3.6_prune_9b-3.6_base	Qwen3-4B
Parameters	8.1B	4.0B
Architecture	qwen35	qwen3
Layers	28	36
Native context	262,144	40,960
Mixture of experts	no	no
Quantizations published	36	29
Smallest quantization	2.35 GiB	1.01 GiB
Q4_K_M	4.76 GiB	2.33 GiB
Licence	apache-2.0	—

KV cache by context

the term that decides long-context viability

Context	MARTHA-LXVII.8B_QWEN-3.5-3.6_prune_9b-3.6_base	Qwen3-4B	Ratio
4,096	0.11 GiB	0.56 GiB	5.14×
8,192	0.22 GiB	1.13 GiB	5.14×
16,384	0.44 GiB	2.25 GiB	5.14×
32,768	0.88 GiB	4.50 GiB	5.14×
65,536	1.75 GiB	9.00 GiB	5.14×
131,072	3.50 GiB	18.00 GiB	5.14×