Model comparison

parakeet-tdt_ctc-110m vs whisper-small.en

At Q4_K_M, parakeet-tdt_ctc-110m is the smaller download — 89,989,600 bytes against 171,553,856.

From the file· summed bytes, KV per layer

Side by side

parakeet-tdt_ctc-110mwhisper-small.en
Parameters114M242M
Architectureparakeetwhisper
Layers12
Native context
Mixture of expertsnono
Quantizations published86
Smallest quantization0.07 GiB0.16 GiB
Q4_K_M0.08 GiB0.16 GiB
Licencecc-by-4.0apache-2.0

KV cache by context

the term that decides long-context viability
Contextparakeet-tdt_ctc-110mwhisper-small.enRatio
4,096
8,192
16,384
32,768
65,536
131,072