Is Falcon3-10B-Instruct or gemma-4-E4B-it smaller?

gemma-4-E4B-it is smaller at Q4_K_M: 4.64 GiB against 5.86 GiB.

Which uses less memory at long context?

gemma-4-E4B-it — its KV cache at 32K context is 9.9x smaller, because of how its attention layers cache.

Model comparison

Falcon3-10B-Instruct vs gemma-4-E4B-it

At Q4_K_M, gemma-4-E4B-it is the smaller download — 4,977,171,584 bytes against 6,287,521,472. At long context the gap widens: gemma-4-E4B-it's KV cache at 32K is 9.9× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Falcon3-10B-Instruct	gemma-4-E4B-it
Parameters	10.3B	8.0B
Architecture	llama	gemma4
Layers	40	42
Native context	32,768	131,072
Mixture of experts	no	no
Quantizations published	43	36
Smallest quantization	2.20 GiB	3.30 GiB
Q4_K_M	5.86 GiB	4.64 GiB
Licence	other	apache-2.0

KV cache by context

the term that decides long-context viability

Context	Falcon3-10B-Instruct	gemma-4-E4B-it	Ratio
4,096	0.63 GiB	0.12 GiB	5.08×
8,192	1.25 GiB	0.18 GiB	7.03×
16,384	2.50 GiB	0.29 GiB	8.71×
32,768	5.00 GiB	0.51 GiB	9.88×
65,536	10.00 GiB	0.94 GiB	10.60×
131,072	20.00 GiB	1.82 GiB	11.00×