Is OpenCoder-1.5B-Instruct or Llama-3.1-8B-Instruct smaller?

OpenCoder-1.5B-Instruct is smaller at Q4_K_M: 1.32 GiB against 4.58 GiB.

Which uses less memory at long context?

Llama-3.1-8B-Instruct — its KV cache at 32K context is 1.6x smaller, because of how its attention layers cache.

Model comparison

OpenCoder-1.5B-Instruct vs Llama-3.1-8B-Instruct

At Q4_K_M, OpenCoder-1.5B-Instruct is the smaller download — 1,417,775,200 bytes against 4,920,739,168. At long context the gap widens: Llama-3.1-8B-Instruct's KV cache at 32K is 1.6× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	OpenCoder-1.5B-Instruct	Llama-3.1-8B-Instruct
Parameters	1.9B	8.0B
Architecture	llama	llama
Layers	24	32
Native context	4,096	131,072
Mixture of experts	no	no
Quantizations published	4	45
Smallest quantization	1.18 GiB	2.02 GiB
Q4_K_M	1.32 GiB	4.58 GiB
Licence	other	llama3.1

KV cache by context

the term that decides long-context viability

Context	OpenCoder-1.5B-Instruct	Llama-3.1-8B-Instruct	Ratio
4,096	0.82 GiB	0.50 GiB	1.64×
8,192	1.64 GiB	1.00 GiB	1.64×
16,384	3.28 GiB	2.00 GiB	1.64×
32,768	6.56 GiB	4.00 GiB	1.64×
65,536	13.13 GiB	8.00 GiB	1.64×
131,072	26.25 GiB	16.00 GiB	1.64×