Is granite-4.0-h-3b-ar or Qwen3-1.7B smaller?

Qwen3-1.7B is smaller at Q4_K_M: 1.03 GiB against 1.92 GiB.

Which uses less memory at long context?

granite-4.0-h-3b-ar — its KV cache at 32K context is 14.0x smaller, because of how its attention layers cache.

Model comparison

granite-4.0-h-3b-ar vs Qwen3-1.7B

At Q4_K_M, Qwen3-1.7B is the smaller download — 1,107,409,472 bytes against 2,058,170,400. At long context the gap widens: granite-4.0-h-3b-ar's KV cache at 32K is 14.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	granite-4.0-h-3b-ar	Qwen3-1.7B
Parameters	3.4B	2.0B
Architecture	granitehybrid	qwen3
Layers	40	28
Native context	131,072	40,960
Mixture of experts	no	no
Quantizations published	36	48
Smallest quantization	0.76 GiB	0.50 GiB
Q4_K_M	1.92 GiB	1.03 GiB
Licence	apache-2.0	—

KV cache by context

the term that decides long-context viability

Context	granite-4.0-h-3b-ar	Qwen3-1.7B	Ratio
4,096	0.03 GiB	0.44 GiB	13.97×
8,192	0.06 GiB	0.88 GiB	13.98×
16,384	0.13 GiB	1.75 GiB	13.99×
32,768	0.25 GiB	3.50 GiB	14.00×
65,536	0.50 GiB	7.00 GiB	14.00×
131,072	1.00 GiB	14.00 GiB	14.00×