Model comparison

Qwen3-Coder-30B-A3B-Instruct vs Kimi-Linear-48B-A3B-Instruct

At Q4_K_M, Qwen3-Coder-30B-A3B-Instruct is the smaller download — 18,556,689,568 bytes against 29,702,758,784. At long context the gap widens: Kimi-Linear-48B-A3B-Instruct's KV cache at 32K is 3.2× 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-InstructKimi-Linear-48B-A3B-Instruct
Parameters30.5B49.1B
Architectureqwen3moekimi-linear
Layers4827
Native context262,144—
Mixture of expertsyes, 128 expertsyes, 256 experts
Quantizations published4638
Smallest quantization7.46 GiB9.77 GiB
Q4_K_M17.28 GiB27.66 GiB
Licenceapache-2.0mit

KV cache by context

the term that decides long-context viability
ContextQwen3-Coder-30B-A3B-InstructKimi-Linear-48B-A3B-InstructRatio
4,0960.38 GiB0.12 GiB3.16×
8,1920.75 GiB0.24 GiB3.16×
16,3841.50 GiB0.47 GiB3.16×
32,7683.00 GiB0.95 GiB3.16×
65,5366.00 GiB1.90 GiB3.16×
131,07212.00 GiB3.80 GiB3.16×