A Blog post by NVIDIA on Hugging Face
NVIDIA Nemotron 3.5 ASR is an open-weights 600M streaming speech model transcribing 40 language-locales with configurable latency.