Model comparison

Qwen3-Coder-30B-A3B-Instruct vs Wizard-Vicuna-13B-Uncensored

At Q4_K_M, Wizard-Vicuna-13B-Uncensored is the smaller download — 7,865,956,256 bytes against 18,556,689,568. At long context the gap widens: Qwen3-Coder-30B-A3B-Instruct's KV cache at 32K is 8.3× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

Qwen3-Coder-30B-A3B-InstructWizard-Vicuna-13B-Uncensored
Parameters30.5B13.0B
Architectureqwen3moellama
Layers4840
Native context262,1442,048
Mixture of expertsyes, 128 expertsno
Quantizations published4625
Smallest quantization7.46 GiB4.52 GiB
Q4_K_M17.28 GiB7.33 GiB
Licenceapache-2.0other

KV cache by context

the term that decides long-context viability
ContextQwen3-Coder-30B-A3B-InstructWizard-Vicuna-13B-UncensoredRatio
4,0960.38 GiB3.13 GiB8.33×
8,1920.75 GiB6.25 GiB8.33×
16,3841.50 GiB12.50 GiB8.33×
32,7683.00 GiB25.00 GiB8.33×
65,5366.00 GiB50.00 GiB8.33×
131,07212.00 GiB100.00 GiB8.33×