Model comparison

nemotron-3.5-asr-streaming-0.6b vs moonshine-streaming-medium

These two publish different quantization sets; the table below has the exact sizes.

From the file· summed bytes, KV per layer

Side by side

nemotron-3.5-asr-streaming-0.6bmoonshine-streaming-medium
Parameters638M266M
Architectureparakeetmoonshine_streaming
Layers14
Native context4,096
Mixture of expertsnono
Quantizations published83
Smallest quantization0.38 GiB0.28 GiB
Q4_K_M0.46 GiB
Licenceothermit

KV cache by context

the term that decides long-context viability
Contextnemotron-3.5-asr-streaming-0.6bmoonshine-streaming-mediumRatio
4,0960.14 GiB
8,1920.27 GiB
16,3840.55 GiB
32,7681.09 GiB
65,5362.19 GiB
131,0724.38 GiB