Is Qwen3-Coder-30B-A3B-Instruct or GLM-4.7-Flash-heretic smaller?

GLM-4.7-Flash-heretic is smaller at Q4_K_M: 17.27 GiB against 17.28 GiB.

Which uses less memory at long context?

GLM-4.7-Flash-heretic — its KV cache at 32K context is 1.8x smaller, because of how its attention layers cache.

Model comparison

Qwen3-Coder-30B-A3B-Instruct vs GLM-4.7-Flash-heretic

At Q4_K_M, GLM-4.7-Flash-heretic is the smaller download — 18,540,437,472 bytes against 18,556,689,568. At long context the gap widens: GLM-4.7-Flash-heretic's KV cache at 32K is 1.8× 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-Instruct	GLM-4.7-Flash-heretic
Parameters	30.5B	29.9B
Architecture	qwen3moe	deepseek2
Layers	48	47
Native context	262,144	202,752
Mixture of experts	yes, 128 experts	yes, 64 experts
Quantizations published	46	26
Smallest quantization	7.46 GiB	7.95 GiB
Q4_K_M	17.28 GiB	17.27 GiB
Licence	apache-2.0	mit

KV cache by context

the term that decides long-context viability

Context	Qwen3-Coder-30B-A3B-Instruct	GLM-4.7-Flash-heretic	Ratio
4,096	0.38 GiB	0.21 GiB	1.82×
8,192	0.75 GiB	0.41 GiB	1.82×
16,384	1.50 GiB	0.83 GiB	1.82×
32,768	3.00 GiB	1.65 GiB	1.82×
65,536	6.00 GiB	3.30 GiB	1.82×
131,072	12.00 GiB	6.61 GiB	1.82×