Is NVIDIA-Nemotron-3-Nano-4B or llama-3-youko-8b smaller?

Both publish quantizations; compare the table for the exact sizes.

Which uses less memory at long context?

llama-3-youko-8b — its KV cache at 32K context is 1.3x smaller, because of how its attention layers cache.

Model comparison

NVIDIA-Nemotron-3-Nano-4B vs llama-3-youko-8b

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: llama-3-youko-8b's KV cache at 32K is 1.3× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	NVIDIA-Nemotron-3-Nano-4B	llama-3-youko-8b
Parameters	4.0B	8.0B
Architecture	nemotron_h	llama
Layers	42	32
Native context	262,144	8,192
Mixture of experts	no	no
Quantizations published	1	2
Smallest quantization	2.64 GiB	5.34 GiB
Q4_K_M	2.64 GiB	—
Licence	other	llama3

KV cache by context

the term that decides long-context viability

Context	NVIDIA-Nemotron-3-Nano-4B	llama-3-youko-8b	Ratio
4,096	0.66 GiB	0.50 GiB	1.31×
8,192	1.31 GiB	1.00 GiB	1.31×
16,384	2.63 GiB	2.00 GiB	1.31×
32,768	5.25 GiB	4.00 GiB	1.31×
65,536	10.50 GiB	8.00 GiB	1.31×
131,072	21.00 GiB	16.00 GiB	1.31×