Is Ministral-3-8B-Instruct-2512-BF16-abliterated or Qwen3-4B smaller?

Qwen3-4B is smaller at Q4_K_M: 2.33 GiB against 4.84 GiB.

Which uses less memory at long context?

Ministral-3-8B-Instruct-2512-BF16-abliterated — its KV cache at 32K context is 1.1x smaller, because of how its attention layers cache.

Model comparison

At Q4_K_M, Qwen3-4B is the smaller download — 2,497,280,256 bytes against 5,198,387,680.

From the file· summed bytes, KV per layer

	Ministral-3-8B-Instruct-2512-BF16-abliterated	Qwen3-4B
Parameters	8.9B	4.0B
Architecture	mistral3	qwen3
Layers	34	36
Native context	262,144	40,960
Mixture of experts	no	no
Quantizations published	36	29
Smallest quantization	1.98 GiB	1.01 GiB
Q4_K_M	4.84 GiB	2.33 GiB
Licence	apache-2.0	—

the term that decides long-context viability

Context	Ministral-3-8B-Instruct-2512-BF16-abliterated	Qwen3-4B	Ratio
4,096	0.53 GiB	0.56 GiB	1.06×
8,192	1.06 GiB	1.13 GiB	1.06×
16,384	2.13 GiB	2.25 GiB	1.06×
32,768	4.25 GiB	4.50 GiB	1.06×
65,536	8.50 GiB	9.00 GiB	1.06×
131,072	17.00 GiB	18.00 GiB	1.06×