Nvidia · Nemotron
Nemotron-3-Diarization
Nvidia's 100M streaming and offline speaker-diarization model for recordings with up to eight speakers.
- Parameters
- 100M
- Input
- Audio
- Output
- Speaker labels
- Access
- Open weights
- Licence
- OpenMDW 1.1
- Released
- Device evidence
- Demonstrated on device
- Last verified
- Formats
- Safetensors, GGUF
- Runtimes
- Transformers, NeMo, NeMo-Speech.cpp
- Platforms
- iOS, Linux, macOS
Argmax has integrated the model into its on-device Pro SDK for Mac and iPhone. Nvidia’s published benchmark throughput uses a desktop GPU rather than mobile hardware.