Model comparison

GRM-Kerlin-3b vs Llama-3.2-1B-Instruct

At Q4_K_M, Llama-3.2-1B-Instruct is the smaller download — 807,690,688 bytes against 2,104,933,344.

From the file· summed bytes, KV per layer

Side by side

GRM-Kerlin-3bLlama-3.2-1B-Instruct
Parameters3.4B1.2B
Architectureqwen2llama
Layers3616
Native context32,768131,072
Mixture of expertsnono
Quantizations published3639
Smallest quantization0.83 GiB0.39 GiB
Q4_K_M1.96 GiB0.75 GiB
Licenceapache-2.0

KV cache by context

the term that decides long-context viability
ContextGRM-Kerlin-3bLlama-3.2-1B-InstructRatio
4,0960.14 GiB0.13 GiB1.13×
8,1920.28 GiB0.25 GiB1.13×
16,3840.56 GiB0.50 GiB1.13×
32,7681.13 GiB1.00 GiB1.13×
65,5362.25 GiB2.00 GiB1.13×
131,0724.50 GiB4.00 GiB1.13×