Model comparison

Slava-Qwen3-14B-Serbian vs Qwen3-Coder-30B-A3B-Instruct

At Q4_K_M, Slava-Qwen3-14B-Serbian is the smaller download — 9,001,753,184 bytes against 18,556,689,568. At long context the gap widens: Qwen3-Coder-30B-A3B-Instruct's KV cache at 32K is 1.7× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

Slava-Qwen3-14B-SerbianQwen3-Coder-30B-A3B-Instruct
Parameters14.8B30.5B
Architectureqwen3qwen3moe
Layers4048
Native context40,960262,144
Mixture of expertsnoyes, 128 experts
Quantizations published3546
Smallest quantization3.33 GiB7.46 GiB
Q4_K_M8.38 GiB17.28 GiB
Licencecc-by-nc-sa-4.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextSlava-Qwen3-14B-SerbianQwen3-Coder-30B-A3B-InstructRatio
4,0960.63 GiB0.38 GiB1.67×
8,1921.25 GiB0.75 GiB1.67×
16,3842.50 GiB1.50 GiB1.67×
32,7685.00 GiB3.00 GiB1.67×
65,53610.00 GiB6.00 GiB1.67×
131,07220.00 GiB12.00 GiB1.67×
Slava-Qwen3-14B-Serbian vs Qwen3-Coder-30B-A3B-Instruct — size, memory and hardware fit — ossmodeldb