Is DeepSeek-V4-Flash or ERNIE-4.5-300B-A47B-PT 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 107.2x smaller, because of how its attention layers cache.

Model comparison

DeepSeek-V4-Flash vs ERNIE-4.5-300B-A47B-PT

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

From the file· summed bytes, KV per layer

Side by side

	DeepSeek-V4-Flash	ERNIE-4.5-300B-A47B-PT
Parameters	291B	300B
Architecture	deepseek4	ernie4_5-moe
Layers	43	54
Native context	1,048,576	131,072
Mixture of experts	yes, 256 experts	no
Quantizations published	12	21
Smallest quantization	76.87 GiB	71.48 GiB
Q4_K_M	—	167.78 GiB
Licence	mit	apache-2.0

KV cache by context

the term that decides long-context viability

Context	DeepSeek-V4-Flash	ERNIE-4.5-300B-A47B-PT	Ratio
4,096	0.06 GiB	0.84 GiB	13.40×
8,192	0.06 GiB	1.69 GiB	26.79×
16,384	0.06 GiB	3.38 GiB	53.58×
32,768	0.06 GiB	6.75 GiB	107.16×
65,536	0.06 GiB	13.50 GiB	214.33×
131,072	0.06 GiB	27.00 GiB	428.65×