Is Trinity-Large-TrueBase 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 42.3x smaller, because of how its attention layers cache.

Model comparison

Trinity-Large-TrueBase 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 42.3× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Trinity-Large-TrueBase	DeepSeek-V4-Flash
Parameters	399B	291B
Architecture	afmoe	deepseek4
Layers	60	43
Native context	8,192	1,048,576
Mixture of experts	yes, 256 experts	yes, 256 experts
Quantizations published	39	12
Smallest quantization	73.49 GiB	76.87 GiB
Q4_K_M	225.04 GiB	—
Licence	other	mit

KV cache by context

the term that decides long-context viability

Context	Trinity-Large-TrueBase	DeepSeek-V4-Flash	Ratio
4,096	0.94 GiB	0.06 GiB	14.88×
8,192	1.26 GiB	0.06 GiB	20.00×
16,384	1.73 GiB	0.06 GiB	27.44×
32,768	2.67 GiB	0.06 GiB	42.33×
65,536	4.54 GiB	0.06 GiB	72.09×
131,072	8.29 GiB	0.06 GiB	131.63×