Is HarmonicHarlequin_v5-20B or Qwen3.8-27B smaller?

Qwen3.8-27B is smaller at Q4_K_M: 15.66 GiB against 18.71 GiB.

Which uses less memory at long context?

Qwen3.8-27B — its KV cache at 32K context is 32.5x smaller, because of how its attention layers cache.

Model comparison

HarmonicHarlequin_v5-20B vs Qwen3.8-27B

At Q4_K_M, Qwen3.8-27B is the smaller download — 16,810,714,336 bytes against 20,087,800,544. At long context the gap widens: Qwen3.8-27B's KV cache at 32K is 32.5× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	HarmonicHarlequin_v5-20B	Qwen3.8-27B
Parameters	33.3B	27.8B
Architecture	llama	qwen35
Layers	104	64
Native context	4,096	262,144
Mixture of experts	no	no
Quantizations published	35	22
Smallest quantization	6.77 GiB	8.39 GiB
Q4_K_M	18.71 GiB	15.66 GiB
Licence	—	apache-2.0

KV cache by context

the term that decides long-context viability

Context	HarmonicHarlequin_v5-20B	Qwen3.8-27B	Ratio
4,096	8.13 GiB	0.25 GiB	32.48×
8,192	16.25 GiB	0.50 GiB	32.49×
16,384	32.50 GiB	1.00 GiB	32.49×
32,768	65.00 GiB	2.00 GiB	32.50×
65,536	130.00 GiB	4.00 GiB	32.50×
131,072	260.00 GiB	8.00 GiB	32.50×