nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B
AI model · https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B
First seen by Oriel on .
AI summary
NVIDIA Nemotron-3.5-Transcribe is an open-weight offline speech and audio model supporting multilingual transcription, translation, speaker attribution, timestamps, and audio question answering.
- The hybrid Mamba-Transformer MoE has 34B total parameters, approximately 4.3B active, and supports BF16, FP8, and NVFP4 variants.
- It processes recordings up to one hour with a 65,536-token context and supports 40 languages, though quality and translation coverage vary by language.
- The model runs on one H100 80 GB through vLLM, while optimized NVIDIA serving reaches 2,238 audio-seconds per second on an RTX PRO 6000 Blackwell GPU.
- Prompts control verbatim or readable transcription, vocabulary hints, speaker tags, translation, timestamps, and audio reasoning; reasoning is recommended only for audio understanding.
- It is intended for developers and researchers building speech interfaces, transcription and translation tools, meeting and media processing, customer-service analytics, and audio-understanding applications.
#speech recognition #audio models #mixture of experts #multilingual ai
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Likes
Total likes reported by the source.
- Hugging Face Models: 3,390, latest observation ; +3,273 over observed period since
Downloads (30d)
Rolling 30-day downloads. A change in this value is not new downloads during the observed period.
- Hugging Face Models: 82,914, latest observation ; +82,576 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Hugging Face Models | nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B | https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B Read discussion |