Is GLM-Z1-9B-0414 or llama-3-youko-8b smaller?

Both publish quantizations; compare the table for the exact sizes.

Which uses less memory at long context?

GLM-Z1-9B-0414 — its KV cache at 32K context is 3.2x smaller, because of how its attention layers cache.

Model comparison

GLM-Z1-9B-0414 vs llama-3-youko-8b

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: GLM-Z1-9B-0414'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

	GLM-Z1-9B-0414	llama-3-youko-8b
Parameters	9.4B	8.0B
Architecture	glm4	llama
Layers	40	32
Native context	32,768	8,192
Mixture of experts	no	no
Quantizations published	47	2
Smallest quantization	2.98 GiB	5.34 GiB
Q4_K_M	5.74 GiB	—
Licence	mit	llama3

KV cache by context

the term that decides long-context viability

Context	GLM-Z1-9B-0414	llama-3-youko-8b	Ratio
4,096	0.16 GiB	0.50 GiB	3.20×
8,192	0.31 GiB	1.00 GiB	3.20×
16,384	0.63 GiB	2.00 GiB	3.20×
32,768	1.25 GiB	4.00 GiB	3.20×
65,536	2.50 GiB	8.00 GiB	3.20×
131,072	5.00 GiB	16.00 GiB	3.20×