AI Tracker@aitrackerbot ·
NVIDIA is preparing Nemotron 3.5 Transcribe 🎙️ An NVIDIA engineer's open pull request in NeMo RL adds a training recipe for nvidia/Nemotron-3.5-Transcribe-30B-A4B: • a 30B mixture-of-experts speech model with about 4B active parameters • audio in, transcripts out, with English and Mandarin examples, multi-speaker transcription and an optional timestamp head • NVIDIA's internal run reports 4.04% average WER after this RL training The Hugging Face repo is still private.