Is EVA-Yi-1.5-9B-32K-V1 or Qwen3-4B smaller?

Qwen3-4B is smaller at Q4_K_M: 2.33 GiB against 4.96 GiB.

Which uses less memory at long context?

EVA-Yi-1.5-9B-32K-V1 — its KV cache at 32K context is 1.5x smaller, because of how its attention layers cache.

Model comparison

EVA-Yi-1.5-9B-32K-V1 vs Qwen3-4B

At Q4_K_M, Qwen3-4B is the smaller download — 2,497,280,256 bytes against 5,328,958,656. At long context the gap widens: EVA-Yi-1.5-9B-32K-V1's KV cache at 32K is 1.5× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	EVA-Yi-1.5-9B-32K-V1	Qwen3-4B
Parameters	8.8B	4.0B
Architecture	llama	qwen3
Layers	48	36
Native context	32,768	40,960
Mixture of experts	no	no
Quantizations published	41	29
Smallest quantization	1.88 GiB	1.01 GiB
Q4_K_M	4.96 GiB	2.33 GiB
Licence	apache-2.0	—

KV cache by context

the term that decides long-context viability

Context	EVA-Yi-1.5-9B-32K-V1	Qwen3-4B	Ratio
4,096	0.38 GiB	0.56 GiB	1.50×
8,192	0.75 GiB	1.13 GiB	1.50×
16,384	1.50 GiB	2.25 GiB	1.50×
32,768	3.00 GiB	4.50 GiB	1.50×
65,536	6.00 GiB	9.00 GiB	1.50×
131,072	12.00 GiB	18.00 GiB	1.50×