NVIDIA releases Nemotron 3 speaker diarization model

NVIDIA released Nemotron 3 Diarization, an open-weight 100M-parameter model that ranks first on VoiceArena's diarization leaderboard with a 14.72% Diarization Error Rate. It supports up to eight speakers in live and recorded conversations, handles overlapping speech, and offers customizable streaming latency.

• Nvidia releases free 100M-parameter model for real-time speaker diarization2026-09-27
• Users snap up RTX 5060 Ti 16GB cards for AI workloads2026-11-08
• 31B video-audio diffusion model compressed to 14B2026-11-07
• US cleared Nvidia H200 chip handler New Matrix despite empty office2026-11-07
• Gary Marcus warns OpenAI could be 'WeWork of AI,' Nvidia may pay2026-11-07