Is command-a-plus-05-2026-bf16 or DeepSeek-V4-Flash 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

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

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

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

KV cache by context

the term that decides long-context viability

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