Model comparison

Hunyuan-A13B-Instruct vs Qwen3-30B-A3B-Thinking-2507

At Q4_K_M, Qwen3-30B-A3B-Thinking-2507 is the smaller download — 18,556,685,824 bytes against 48,782,258,016. At long context the gap widens: Qwen3-30B-A3B-Thinking-2507's KV cache at 32K is 1.3× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

Hunyuan-A13B-InstructQwen3-30B-A3B-Thinking-2507
Parameters80.4B30.5B
Architecturehunyuan-moeqwen3moe
Layers3248
Native context32,768262,144
Mixture of expertsyes, 64 expertsyes, 128 experts
Quantizations published7051
Smallest quantization15.18 GiB7.05 GiB
Q4_K_M45.43 GiB17.28 GiB
Licenceotherapache-2.0

KV cache by context

the term that decides long-context viability
ContextHunyuan-A13B-InstructQwen3-30B-A3B-Thinking-2507Ratio
4,0960.50 GiB0.38 GiB1.33×
8,1921.00 GiB0.75 GiB1.33×
16,3842.00 GiB1.50 GiB1.33×
32,7684.00 GiB3.00 GiB1.33×
65,5368.00 GiB6.00 GiB1.33×
131,07216.00 GiB12.00 GiB1.33×