Is Gemma-4-31B-Isometry-Fabled-Persona or llama-3-youko-8b smaller?

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

Which uses less memory at long context?

llama-3-youko-8b — its KV cache at 32K context is 1.5x smaller, because of how its attention layers cache.

Model comparison

Gemma-4-31B-Isometry-Fabled-Persona vs llama-3-youko-8b

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: llama-3-youko-8b's KV cache at 32K is 1.5× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Gemma-4-31B-Isometry-Fabled-Persona	llama-3-youko-8b
Parameters	31.3B	8.0B
Architecture	gemma4	llama
Layers	60	32
Native context	262,144	8,192
Mixture of experts	no	no
Quantizations published	34	2
Smallest quantization	6.67 GiB	5.34 GiB
Q4_K_M	17.40 GiB	—
Licence	apache-2.0	llama3

KV cache by context

the term that decides long-context viability

Context	Gemma-4-31B-Isometry-Fabled-Persona	llama-3-youko-8b	Ratio
4,096	1.80 GiB	0.50 GiB	3.59×
8,192	2.42 GiB	1.00 GiB	2.42×
16,384	3.67 GiB	2.00 GiB	1.84×
32,768	6.17 GiB	4.00 GiB	1.54×
65,536	11.17 GiB	8.00 GiB	1.40×
131,072	21.17 GiB	16.00 GiB	1.32×