Is gemma-4-31B-it-Mystery-Fine-Tune-HERETIC-UNCENSORED-Thinking or Qwen3.5-9B smaller?

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

Which uses less memory at long context?

Qwen3.5-9B — its KV cache at 32K context is 6.2x smaller, because of how its attention layers cache.

Model comparison

gemma-4-31B-it-Mystery-Fine-Tune-HERETIC-UNCENSORED-Thinking vs Qwen3.5-9B

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: Qwen3.5-9B's KV cache at 32K is 6.2× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	gemma-4-31B-it-Mystery-Fine-Tune-HERETIC-UNCENSORED-Thinking	Qwen3.5-9B
Parameters	31.3B	9.7B
Architecture	gemma4	qwen35
Layers	60	32
Native context	262,144	262,144
Mixture of experts	no	no
Quantizations published	27	34
Smallest quantization	6.66 GiB	2.97 GiB
Q4_K_M	—	5.24 GiB
Licence	apache-2.0	apache-2.0

KV cache by context

the term that decides long-context viability

Context	gemma-4-31B-it-Mystery-Fine-Tune-HERETIC-UNCENSORED-Thinking	Qwen3.5-9B	Ratio
4,096	1.80 GiB	0.13 GiB	14.36×
8,192	2.42 GiB	0.25 GiB	9.68×
16,384	3.67 GiB	0.50 GiB	7.34×
32,768	6.17 GiB	1.00 GiB	6.17×
65,536	11.17 GiB	2.00 GiB	5.59×
131,072	21.17 GiB	4.00 GiB	5.29×