Model comparison

dolphin-2.2.1-mistral-7b 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 4,368,450,304.

From the file· summed bytes, KV per layer

Side by side

dolphin-2.2.1-mistral-7bHyperCLOVAX-SEED-Text-Instruct-1.5B
Parameters7.2B1.6B
Architecturellamallama
Layers32
Native context32,768
Mixture of expertsnono
Quantizations published461
Smallest quantization1.50 GiB1.06 GiB
Q4_K_M4.07 GiB1.06 GiB
Licenceapache-2.0other

KV cache by context

the term that decides long-context viability
Contextdolphin-2.2.1-mistral-7bHyperCLOVAX-SEED-Text-Instruct-1.5BRatio
4,096
8,192
16,384
32,768
65,536
131,072