Model comparison

magnum-v2-4b vs Llama-3.1-8B-Instruct

At Q4_K_M, magnum-v2-4b is the smaller download — 2,778,283,488 bytes against 4,920,739,168.

From the file· summed bytes, KV per layer

Side by side

magnum-v2-4bLlama-3.1-8B-Instruct
Parameters4.5B8.0B
Architecturellamallama
Layers3232
Native context131,072131,072
Mixture of expertsnono
Quantizations published4145
Smallest quantization1.13 GiB2.02 GiB
Q4_K_M2.59 GiB4.58 GiB
Licenceapache-2.0llama3.1

KV cache by context

the term that decides long-context viability
Contextmagnum-v2-4bLlama-3.1-8B-InstructRatio
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×