Is nemotron-speech-streaming-en-0.6b or parakeet-tdt-0.6b-v3 smaller?

nemotron-speech-streaming-en-0.6b is smaller at Q4_K_M: 0.44 GiB against 0.45 GiB.

Which uses less memory at long context?

We cannot resolve the KV layout for one of these models, so we do not publish a comparison.

Model comparison

At Q4_K_M, nemotron-speech-streaming-en-0.6b is the smaller download — 475,436,032 bytes against 485,425,504.

From the file· summed bytes, KV per layer

	nemotron-speech-streaming-en-0.6b	parakeet-tdt-0.6b-v3
Parameters	618M	627M
Architecture	parakeet	parakeet
Layers	—	—
Native context	—	—
Mixture of experts	no	no
Quantizations published	6	10
Smallest quantization	0.44 GiB	0.39 GiB
Q4_K_M	0.44 GiB	0.45 GiB
Licence	other	cc-by-4.0

the term that decides long-context viability

Context	nemotron-speech-streaming-en-0.6b	parakeet-tdt-0.6b-v3	Ratio
4,096	—	—	—
8,192	—	—	—
16,384	—	—	—
32,768	—	—	—
65,536	—	—	—
131,072	—	—	—