Is DeepSeek-V4-Flash or grok-2 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 127.0x smaller, because of how its attention layers cache.

Model comparison

DeepSeek-V4-Flash vs grok-2

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 127.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	DeepSeek-V4-Flash	grok-2
Parameters	291B	270B
Architecture	deepseek4	grok
Layers	43	64
Native context	1,048,576	131,072
Mixture of experts	yes, 256 experts	yes, 8 experts
Quantizations published	12	66
Smallest quantization	76.87 GiB	57.16 GiB
Q4_K_M	—	152.79 GiB
Licence	mit	other

KV cache by context

the term that decides long-context viability

Context	DeepSeek-V4-Flash	grok-2	Ratio
4,096	0.06 GiB	1.00 GiB	15.88×
8,192	0.06 GiB	2.00 GiB	31.75×
16,384	0.06 GiB	4.00 GiB	63.50×
32,768	0.06 GiB	8.00 GiB	127.01×
65,536	0.06 GiB	16.00 GiB	254.02×
131,072	0.06 GiB	32.00 GiB	508.03×