Is DeepSeek-V4-Flash or command-a-plus-05-2026-bf16 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 22.6x smaller, because of how its attention layers cache.

Model comparison

DeepSeek-V4-Flash vs command-a-plus-05-2026-bf16

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

From the file· summed bytes, KV per layer

Side by side

	DeepSeek-V4-Flash	command-a-plus-05-2026-bf16
Parameters	291B	219B
Architecture	deepseek4	cohere2moe
Layers	43	32
Native context	1,048,576	200,000
Mixture of experts	yes, 256 experts	yes, 128 experts
Quantizations published	12	25
Smallest quantization	76.87 GiB	45.87 GiB
Q4_K_M	—	125.81 GiB
Licence	mit	apache-2.0

KV cache by context

the term that decides long-context viability

Context	DeepSeek-V4-Flash	command-a-plus-05-2026-bf16	Ratio
4,096	0.06 GiB	0.50 GiB	7.94×
8,192	0.06 GiB	0.67 GiB	10.67×
16,384	0.06 GiB	0.92 GiB	14.64×
32,768	0.06 GiB	1.42 GiB	22.57×
65,536	0.06 GiB	2.42 GiB	38.45×
131,072	0.06 GiB	4.42 GiB	70.20×