Is Qwen-3.5-2b-roleplay-tuned or Qwen3-8B smaller?

Qwen-3.5-2b-roleplay-tuned is smaller at Q4_K_M: 1.22 GiB against 4.68 GiB.

Which uses less memory at long context?

Qwen-3.5-2b-roleplay-tuned — its KV cache at 32K context is 12.0x smaller, because of how its attention layers cache.

Model comparison

Qwen-3.5-2b-roleplay-tuned vs Qwen3-8B

At Q4_K_M, Qwen-3.5-2b-roleplay-tuned is the smaller download — 1,312,165,792 bytes against 5,027,783,488. At long context the gap widens: Qwen-3.5-2b-roleplay-tuned's KV cache at 32K is 12.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Qwen-3.5-2b-roleplay-tuned	Qwen3-8B
Parameters	2.3B	8.2B
Architecture	qwen35	qwen3
Layers	24	36
Native context	262,144	40,960
Mixture of experts	no	no
Quantizations published	27	51
Smallest quantization	0.92 GiB	2.12 GiB
Q4_K_M	1.22 GiB	4.68 GiB
Licence	apache-2.0	apache-2.0

KV cache by context

the term that decides long-context viability

Context	Qwen-3.5-2b-roleplay-tuned	Qwen3-8B	Ratio
4,096	0.05 GiB	0.56 GiB	11.99×
8,192	0.09 GiB	1.13 GiB	12.00×
16,384	0.19 GiB	2.25 GiB	12.00×
32,768	0.38 GiB	4.50 GiB	12.00×
65,536	0.75 GiB	9.00 GiB	12.00×
131,072	1.50 GiB	18.00 GiB	12.00×