Is HunyuanOCR or Llama-3.2-1B-Instruct smaller?

HunyuanOCR is smaller at Q4_K_M: 0.33 GiB against 0.75 GiB.

Which uses less memory at long context?

HunyuanOCR — its KV cache at 32K context is 1.2x smaller, because of how its attention layers cache.

Model comparison

HunyuanOCR vs Llama-3.2-1B-Instruct

At Q4_K_M, HunyuanOCR is the smaller download — 354,966,656 bytes against 807,690,688. At long context the gap widens: HunyuanOCR's KV cache at 32K is 1.2× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	HunyuanOCR	Llama-3.2-1B-Instruct
Parameters	1.1B	1.2B
Architecture	hunyuan_vl	llama
Layers	24	16
Native context	131,072	131,072
Mixture of experts	yes, 1 experts	no
Quantizations published	25	39
Smallest quantization	0.24 GiB	0.39 GiB
Q4_K_M	0.33 GiB	0.75 GiB
Licence	—	—

KV cache by context

the term that decides long-context viability

Context	HunyuanOCR	Llama-3.2-1B-Instruct	Ratio
4,096	0.11 GiB	0.13 GiB	1.19×
8,192	0.21 GiB	0.25 GiB	1.19×
16,384	0.42 GiB	0.50 GiB	1.19×
32,768	0.84 GiB	1.00 GiB	1.19×
65,536	1.69 GiB	2.00 GiB	1.19×
131,072	3.38 GiB	4.00 GiB	1.19×