Is Noema-2B or Llama-3.1-8B-Instruct smaller?

Noema-2B is smaller at Q4_K_M: 1.19 GiB against 4.58 GiB.

Which uses less memory at long context?

Noema-2B — its KV cache at 32K context is 10.7x smaller, because of how its attention layers cache.

Model comparison

Noema-2B vs Llama-3.1-8B-Instruct

At Q4_K_M, Noema-2B is the smaller download — 1,274,396,096 bytes against 4,920,739,168. At long context the gap widens: Noema-2B's KV cache at 32K is 10.7× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Noema-2B	Llama-3.1-8B-Instruct
Parameters	1.9B	8.0B
Architecture	qwen35	llama
Layers	24	32
Native context	262,144	131,072
Mixture of experts	no	no
Quantizations published	31	45
Smallest quantization	0.90 GiB	2.02 GiB
Q4_K_M	1.19 GiB	4.58 GiB
Licence	apache-2.0	llama3.1

KV cache by context

the term that decides long-context viability

Context	Noema-2B	Llama-3.1-8B-Instruct	Ratio
4,096	0.05 GiB	0.50 GiB	10.66×
8,192	0.09 GiB	1.00 GiB	10.66×
16,384	0.19 GiB	2.00 GiB	10.66×
32,768	0.38 GiB	4.00 GiB	10.67×
65,536	0.75 GiB	8.00 GiB	10.67×
131,072	1.50 GiB	16.00 GiB	10.67×