Is SuperHY3-abliterated-NVFP4 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 158.8x smaller, because of how its attention layers cache.

Model comparison

SuperHY3-abliterated-NVFP4 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 158.8× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	SuperHY3-abliterated-NVFP4	DeepSeek-V4-Flash
Parameters	172B	291B
Architecture	hy-v3	deepseek4
Layers	80	43
Native context	262,144	1,048,576
Mixture of experts	yes, 192 experts	yes, 256 experts
Quantizations published	1	12
Smallest quantization	95.10 GiB	76.87 GiB
Q4_K_M	—	—
Licence	apache-2.0	mit

KV cache by context

the term that decides long-context viability

Context	SuperHY3-abliterated-NVFP4	DeepSeek-V4-Flash	Ratio
4,096	1.25 GiB	0.06 GiB	19.84×
8,192	2.50 GiB	0.06 GiB	39.69×
16,384	5.00 GiB	0.06 GiB	79.38×
32,768	10.00 GiB	0.06 GiB	158.76×
65,536	20.00 GiB	0.06 GiB	317.52×
131,072	40.00 GiB	0.06 GiB	635.04×