Is DeepSeek-V4-Flash or Hermes-4-405B smaller?

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

Which uses less memory at long context?

DeepSeek-V4-Flash — its KV cache at 32K context is 250.0x smaller, because of how its attention layers cache.

Model comparison

DeepSeek-V4-Flash vs Hermes-4-405B

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: DeepSeek-V4-Flash's KV cache at 32K is 250.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	DeepSeek-V4-Flash	Hermes-4-405B
Parameters	291B	406B
Architecture	deepseek4	llama
Layers	43	126
Native context	1,048,576	131,072
Mixture of experts	yes, 256 experts	no
Quantizations published	12	16
Smallest quantization	76.87 GiB	81.23 GiB
Q4_K_M	—	226.38 GiB
Licence	mit	llama3

KV cache by context

the term that decides long-context viability

Context	DeepSeek-V4-Flash	Hermes-4-405B	Ratio
4,096	0.06 GiB	1.97 GiB	31.26×
8,192	0.06 GiB	3.94 GiB	62.51×
16,384	0.06 GiB	7.88 GiB	125.02×
32,768	0.06 GiB	15.75 GiB	250.05×
65,536	0.06 GiB	31.50 GiB	500.09×
131,072	0.06 GiB	63.00 GiB	1000.19×