Model comparison

NVIDIA-Nemotron-3-Nano-4B-BF16 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 2,837,072,896.

From the file· summed bytes, KV per layer

Side by side

NVIDIA-Nemotron-3-Nano-4B-BF16HyperCLOVAX-SEED-Text-Instruct-1.5B
Parameters4.0B1.6B
Architecturenemotron_hllama
Layers42
Native context262,144
Mixture of expertsnono
Quantizations published421
Smallest quantization2.03 GiB1.06 GiB
Q4_K_M2.64 GiB1.06 GiB
Licenceotherother

KV cache by context

the term that decides long-context viability
ContextNVIDIA-Nemotron-3-Nano-4B-BF16HyperCLOVAX-SEED-Text-Instruct-1.5BRatio
4,0960.66 GiB
8,1921.31 GiB
16,3842.63 GiB
32,7685.25 GiB
65,53610.50 GiB
131,07221.00 GiB