Model comparison

Apriel-1.6-15b-Thinker vs Qwythos-9B-Claude-Mythos-5-1M

At Q4_K_M, Apriel-1.6-15b-Thinker is the smaller download — 8,785,478,144 bytes against 11,516,776,704. At long context the gap widens: Qwythos-9B-Claude-Mythos-5-1M's KV cache at 32K is 6.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

Apriel-1.6-15b-ThinkerQwythos-9B-Claude-Mythos-5-1M
Parameters14.9B9.4B
Architecturellamaqwen35
Layers4832
Native context262,4001,048,576
Mixture of expertsnono
Quantizations published6016
Smallest quantization3.22 GiB5.38 GiB
Q4_K_M8.18 GiB10.73 GiB
Licencemitapache-2.0

KV cache by context

the term that decides long-context viability
ContextApriel-1.6-15b-ThinkerQwythos-9B-Claude-Mythos-5-1MRatio
4,0960.75 GiB0.13 GiB6.00×
8,1921.50 GiB0.25 GiB6.00×
16,3843.00 GiB0.50 GiB6.00×
32,7686.00 GiB1.00 GiB6.00×
65,53612.00 GiB2.00 GiB6.00×
131,07224.00 GiB4.00 GiB6.00×