Model comparison

Llama-3.2-3B-bnb-4bit vs HyperCLOVAX-SEED-Text-Instruct-1.5B

At Q4_K_M, HyperCLOVAX-SEED-Text-Instruct-1.5B is the smaller download — 1,133,974,368 bytes against 2,019,373,344.

From the file· summed bytes, KV per layer

Side by side

Llama-3.2-3B-bnb-4bitHyperCLOVAX-SEED-Text-Instruct-1.5B
Parameters3.3B1.6B
Architecturellamallama
Layers28
Native context131,072
Mixture of expertsnono
Quantizations published11
Smallest quantization1.88 GiB1.06 GiB
Q4_K_M1.88 GiB1.06 GiB
Licenceapache-2.0other

KV cache by context

the term that decides long-context viability
ContextLlama-3.2-3B-bnb-4bitHyperCLOVAX-SEED-Text-Instruct-1.5BRatio
4,0960.44 GiB
8,1920.88 GiB
16,3841.75 GiB
32,7683.50 GiB
65,5367.00 GiB
131,07214.00 GiB