Is bitnet-b1.58-2B-4T or tinygemma3_cifar smaller?

Both publish quantizations; compare the table for the exact sizes.

Which uses less memory at long context?

tinygemma3_cifar — its KV cache at 32K context is 18.9x smaller, because of how its attention layers cache.

Model comparison

bitnet-b1.58-2B-4T vs tinygemma3_cifar

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: tinygemma3_cifar's KV cache at 32K is 18.9× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	bitnet-b1.58-2B-4T	tinygemma3_cifar
Parameters	850M	39M
Architecture	—	gemma3
Layers	30	8
Native context	4,096	131,072
Mixture of experts	no	no
Quantizations published	2	1
Smallest quantization	0.10 GiB	0.04 GiB
Q4_K_M	—	—
Licence	mit	wtfpl

KV cache by context

the term that decides long-context viability

Context	bitnet-b1.58-2B-4T	tinygemma3_cifar	Ratio
4,096	0.29 GiB	0.06 GiB	4.69×
8,192	0.59 GiB	0.08 GiB	7.59×
16,384	1.17 GiB	0.09 GiB	12.63×
32,768	2.34 GiB	0.12 GiB	18.90×
65,536	4.69 GiB	0.19 GiB	25.13×
131,072	9.38 GiB	0.31 GiB	30.09×