Model comparison

Mistral-7B-v0.2 vs llama-3-youko-8b

These two publish different quantization sets; the table below has the exact sizes.

From the file· summed bytes, KV per layer

Side by side

Mistral-7B-v0.2llama-3-youko-8b
Parameters7.2B8.0B
Architecturellamallama
Layers3232
Native context32,7688,192
Mixture of expertsnono
Quantizations published322
Smallest quantization2.53 GiB5.34 GiB
Q4_K_M4.07 GiB
Licenceapache-2.0llama3

KV cache by context

the term that decides long-context viability
ContextMistral-7B-v0.2llama-3-youko-8bRatio
4,0960.50 GiB0.50 GiB1.00×
8,1921.00 GiB1.00 GiB1.00×
16,3842.00 GiB2.00 GiB1.00×
32,7684.00 GiB4.00 GiB1.00×
65,5368.00 GiB8.00 GiB1.00×
131,07216.00 GiB16.00 GiB1.00×