Is DeepSeek-Coder-V2-Instruct-0724 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 33.5x smaller, because of how its attention layers cache.

Model comparison

DeepSeek-Coder-V2-Instruct-0724 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 33.5× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	DeepSeek-Coder-V2-Instruct-0724	DeepSeek-V4-Flash
Parameters	236B	291B
Architecture	deepseek2	deepseek4
Layers	60	43
Native context	163,840	1,048,576
Mixture of experts	yes, 160 experts	yes, 256 experts
Quantizations published	44	12
Smallest quantization	44.14 GiB	76.87 GiB
Q4_K_M	132.67 GiB	—
Licence	other	mit

KV cache by context

the term that decides long-context viability

Context	DeepSeek-Coder-V2-Instruct-0724	DeepSeek-V4-Flash	Ratio
4,096	0.26 GiB	0.06 GiB	4.19×
8,192	0.53 GiB	0.06 GiB	8.37×
16,384	1.05 GiB	0.06 GiB	16.74×
32,768	2.11 GiB	0.06 GiB	33.49×
65,536	4.22 GiB	0.06 GiB	66.98×
131,072	8.44 GiB	0.06 GiB	133.95×