Model comparison

Qwen3-Coder-30B-A3B-Instruct vs MN-12B-Lyra-v4

At Q4_K_M, MN-12B-Lyra-v4 is the smaller download — 7,477,207,776 bytes against 18,556,689,568.

From the file· summed bytes, KV per layer

Side by side

Qwen3-Coder-30B-A3B-InstructMN-12B-Lyra-v4
Parameters30.5B12.2B
Architectureqwen3moellama
Layers48
Native context262,144
Mixture of expertsyes, 128 expertsno
Quantizations published4633
Smallest quantization7.46 GiB4.13 GiB
Q4_K_M17.28 GiB6.96 GiB
Licenceapache-2.0cc-by-nc-4.0

KV cache by context

the term that decides long-context viability
ContextQwen3-Coder-30B-A3B-InstructMN-12B-Lyra-v4Ratio
4,0960.38 GiB
8,1920.75 GiB
16,3841.50 GiB
32,7683.00 GiB
65,5366.00 GiB
131,07212.00 GiB