NVIDIANemotron 3.5
Nemotron 3.5 Transcribe 30B-A4B
Transcribe is an offline speech and audio model for 40 languages. It accepts recordings, instructions and optional context, then returns text.
What the release provides
The card supports recordings up to one hour and a 65,536-token context. About 4.3B parameters are active, including the audio encoder. Tasks include transcription, translation, speaker attribution, timestamps and audio questions.Nemotron 3.5 Transcribe 30B-A4B official model card
How to use it well
Practical guidance. Choose this model for recorded material when the task needs more than a plain transcript. Keep transcription and interpretation in separate output fields. A summary or answer about audio should remain traceable to the recording and its timestamps.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B |
| Release license | OpenMDW 1.1Nemotron 3.5 Transcribe 30B-A4B official model card |