Is WizardCoder-Python-13B-V1.0 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 6.3x smaller, because of how its attention layers cache.

Model comparison

WizardCoder-Python-13B-V1.0 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 6.3× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	WizardCoder-Python-13B-V1.0	llama-3-youko-8b
Parameters	13.0B	8.0B
Architecture	llama	llama
Layers	40	32
Native context	16,384	8,192
Mixture of experts	no	no
Quantizations published	47	2
Smallest quantization	2.70 GiB	5.34 GiB
Q4_K_M	7.33 GiB	—
Licence	llama2	llama3

KV cache by context

the term that decides long-context viability

Context	WizardCoder-Python-13B-V1.0	llama-3-youko-8b	Ratio
4,096	3.13 GiB	0.50 GiB	6.25×
8,192	6.25 GiB	1.00 GiB	6.25×
16,384	12.50 GiB	2.00 GiB	6.25×
32,768	25.00 GiB	4.00 GiB	6.25×
65,536	50.00 GiB	8.00 GiB	6.25×
131,072	100.00 GiB	16.00 GiB	6.25×