Open Source AI Pulse

Nvidia Expands Open ASR with Nemotron 3.5 Transcribe

Nvidia Expands Open ASR with Nemotron 3.5 Transcribe

Recent coverage describes Nemotron 3.5 Transcribe as an open ASR model supporting 40 languages and combining transcription, translation, speaker attribution, and audio Q&A; reports identify a 30B MoE with 4B active parameters, Granary v2, and fine-tuning tools. Reported accuracy, speed, and medical-domain results remain largely vendor-reported, so builders should verify the model card, license, and performance on representative recordings.

Sources (3)
Updated Oct 28, 2026