Is ultravox-v0_5-llama-3_2-1b or gemma-3-1b-it smaller?

gemma-3-1b-it is smaller at Q4_K_M: 0.75 GiB against 0.75 GiB.

Which uses less memory at long context?

gemma-3-1b-it — its KV cache at 32K context is 54.6x smaller, because of how its attention layers cache.

Model comparison

ultravox-v0_5-llama-3_2-1b vs gemma-3-1b-it

At Q4_K_M, gemma-3-1b-it is the smaller download — 806,058,240 bytes against 807,694,464. At long context the gap widens: gemma-3-1b-it's KV cache at 32K is 54.6× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	ultravox-v0_5-llama-3_2-1b	gemma-3-1b-it
Parameters	683M	1000M
Architecture	llama	gemma3
Layers	16	26
Native context	—	32,768
Mixture of experts	no	no
Quantizations published	2	28
Smallest quantization	0.75 GiB	0.52 GiB
Q4_K_M	0.75 GiB	0.75 GiB
Licence	mit	gemma

KV cache by context

the term that decides long-context viability

Context	ultravox-v0_5-llama-3_2-1b	gemma-3-1b-it	Ratio
4,096	1.00 GiB	0.04 GiB	26.95×
8,192	2.00 GiB	0.05 GiB	37.93×
16,384	4.00 GiB	0.08 GiB	47.63×
32,768	8.00 GiB	0.15 GiB	54.61×
65,536	16.00 GiB	0.27 GiB	58.94×
131,072	32.00 GiB	0.52 GiB	61.36×