Is Chocolatine-3B-Instruct-DPO-Revised or Qwen3-4B smaller?

Chocolatine-3B-Instruct-DPO-Revised is smaller at Q4_K_M: 2.23 GiB against 2.33 GiB.

Which uses less memory at long context?

Qwen3-4B — its KV cache at 32K context is 2.7x smaller, because of how its attention layers cache.

Model comparison

Chocolatine-3B-Instruct-DPO-Revised vs Qwen3-4B

At Q4_K_M, Chocolatine-3B-Instruct-DPO-Revised is the smaller download — 2,393,231,840 bytes against 2,497,280,256. At long context the gap widens: Qwen3-4B's KV cache at 32K is 2.7× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	Chocolatine-3B-Instruct-DPO-Revised	Qwen3-4B
Parameters	3.8B	4.0B
Architecture	phi3	qwen3
Layers	32	36
Native context	4,096	40,960
Mixture of experts	no	no
Quantizations published	1	29
Smallest quantization	2.23 GiB	1.01 GiB
Q4_K_M	2.23 GiB	2.33 GiB
Licence	mit	—

KV cache by context

the term that decides long-context viability

Context	Chocolatine-3B-Instruct-DPO-Revised	Qwen3-4B	Ratio
4,096	1.50 GiB	0.56 GiB	2.67×
8,192	3.00 GiB	1.13 GiB	2.67×
16,384	6.00 GiB	2.25 GiB	2.67×
32,768	12.00 GiB	4.50 GiB	2.67×
65,536	24.00 GiB	9.00 GiB	2.67×
131,072	48.00 GiB	18.00 GiB	2.67×