Model comparison

Magistral-Small-2509-Vision vs Qwythos-9B-Claude-Mythos-5-1M

At Q4_K_M, Qwythos-9B-Claude-Mythos-5-1M is the smaller download — 11,516,776,704 bytes against 12,621,946,560. At long context the gap widens: Qwythos-9B-Claude-Mythos-5-1M's KV cache at 32K is 5.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

Magistral-Small-2509-VisionQwythos-9B-Claude-Mythos-5-1M
Parameters24.0B9.4B
Architecturellamaqwen35
Layers4032
Native context131,0721,048,576
Mixture of expertsnono
Quantizations published1616
Smallest quantization8.09 GiB5.38 GiB
Q4_K_M11.76 GiB10.73 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextMagistral-Small-2509-VisionQwythos-9B-Claude-Mythos-5-1MRatio
4,0960.63 GiB0.13 GiB5.00×
8,1921.25 GiB0.25 GiB5.00×
16,3842.50 GiB0.50 GiB5.00×
32,7685.00 GiB1.00 GiB5.00×
65,53610.00 GiB2.00 GiB5.00×
131,07220.00 GiB4.00 GiB5.00×