Model comparison

granite-20b-code-base-8k vs Qwen3-30B-A3B-Thinking-2507

At Q4_K_M, granite-20b-code-base-8k is the smaller download — 12,820,207,360 bytes against 18,556,685,824.

From the file· summed bytes, KV per layer

Side by side

granite-20b-code-base-8kQwen3-30B-A3B-Thinking-2507
Parameters20.1B30.5B
Architecturestarcoderqwen3moe
Layers5248
Native context262,144
Mixture of expertsnoyes, 128 experts
Quantizations published3651
Smallest quantization4.21 GiB7.05 GiB
Q4_K_M11.94 GiB17.28 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
Contextgranite-20b-code-base-8kQwen3-30B-A3B-Thinking-2507Ratio
4,0960.38 GiB
8,1920.75 GiB
16,3841.50 GiB
32,7683.00 GiB
65,5366.00 GiB
131,07212.00 GiB
granite-20b-code-base-8k vs Qwen3-30B-A3B-Thinking-2507 — size, memory and hardware fit — ossmodeldb