Is Qwen3.5-4B-SOMPOA-heresy or Llama-3.1-8B-Instruct smaller?

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

Which uses less memory at long context?

Qwen3.5-4B-SOMPOA-heresy — its KV cache at 32K context is 4.0x smaller, because of how its attention layers cache.

Model comparison

Qwen3.5-4B-SOMPOA-heresy vs Llama-3.1-8B-Instruct

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: Qwen3.5-4B-SOMPOA-heresy's KV cache at 32K is 4.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Qwen3.5-4B-SOMPOA-heresy	Llama-3.1-8B-Instruct
Parameters	4.5B	8.0B
Architecture	qwen35	llama
Layers	32	32
Native context	262,144	131,072
Mixture of experts	no	no
Quantizations published	24	45
Smallest quantization	1.09 GiB	2.02 GiB
Q4_K_M	—	4.58 GiB
Licence	apache-2.0	llama3.1

KV cache by context

the term that decides long-context viability

Context	Qwen3.5-4B-SOMPOA-heresy	Llama-3.1-8B-Instruct	Ratio
4,096	0.13 GiB	0.50 GiB	4.00×
8,192	0.25 GiB	1.00 GiB	4.00×
16,384	0.50 GiB	2.00 GiB	4.00×
32,768	1.00 GiB	4.00 GiB	4.00×
65,536	2.00 GiB	8.00 GiB	4.00×
131,072	4.00 GiB	16.00 GiB	4.00×