Model comparison

ACE-Step-v1-3.5B vs Llama-3.2-1B-Instruct

At Q4_K_M, Llama-3.2-1B-Instruct is the smaller download — 807,690,688 bytes against 6,149,775,712.

From the file· summed bytes, KV per layer

Side by side

ACE-Step-v1-3.5BLlama-3.2-1B-Instruct
Parameters3.3B1.2B
Architecturepigllama
Layers16
Native context131,072
Mixture of expertsnono
Quantizations published1939
Smallest quantization0.42 GiB0.39 GiB
Q4_K_M5.73 GiB0.75 GiB
Licenceapache-2.0

KV cache by context

the term that decides long-context viability
ContextACE-Step-v1-3.5BLlama-3.2-1B-InstructRatio
4,0960.13 GiB
8,1920.25 GiB
16,3840.50 GiB
32,7681.00 GiB
65,5362.00 GiB
131,0724.00 GiB