Is granite-3.1-1b-a400m-instruct or HyperCLOVAX-SEED-Text-Instruct-1.5B smaller?

granite-3.1-1b-a400m-instruct is smaller at Q4_K_M: 0.77 GiB against 1.06 GiB.

Which uses less memory at long context?

We cannot resolve the KV layout for one of these models, so we do not publish a comparison.

Model comparison

At Q4_K_M, granite-3.1-1b-a400m-instruct is the smaller download — 821,847,360 bytes against 1,133,974,368.

From the file· summed bytes, KV per layer

	granite-3.1-1b-a400m-instruct	HyperCLOVAX-SEED-Text-Instruct-1.5B
Parameters	1.3B	1.6B
Architecture	granitemoe	llama
Layers	24	—
Native context	131,072	—
Mixture of experts	yes, 32 experts	no
Quantizations published	20	1
Smallest quantization	0.48 GiB	1.06 GiB
Q4_K_M	0.77 GiB	1.06 GiB
Licence	apache-2.0	other

the term that decides long-context viability

Context	granite-3.1-1b-a400m-instruct	HyperCLOVAX-SEED-Text-Instruct-1.5B	Ratio
4,096	0.19 GiB	—	—
8,192	0.38 GiB	—	—
16,384	0.75 GiB	—	—
32,768	1.50 GiB	—	—
65,536	3.00 GiB	—	—
131,072	6.00 GiB	—	—