Key Info

NVIDIA announced Nemotron 3.5 Transcribe, an open-weight, promptable 30B Mixture-of-Experts model with roughly 4B active parameters. It combines transcription in 40 languages, speaker attribution, translation, contextual prompting and audio understanding into a single checkpoint.

Highlights

  • Open-weight 30B MoE architecture with about 4B active parameters
  • Transcription support for 40 languages
  • Speaker attribution, including tracking who said what in overlapping speech
  • Translation, contextual prompting and audio understanding in one model
  • Positioned as a promptable model rather than a fixed transcription pipeline